Papers by various scientists on mathematics and computer science, presented at The International Symposium
Research and Education in Innovation Era,
5th Edition,
Arad, Romania, November 5th – 7th, 2014
(ISREIE 2014).

© All Rights Reserved

73 tayangan

Papers by various scientists on mathematics and computer science, presented at The International Symposium
Research and Education in Innovation Era,
5th Edition,
Arad, Romania, November 5th – 7th, 2014
(ISREIE 2014).

© All Rights Reserved

- Pseudo Lattice Graphs and their Applications to Fuzzy and Neutrosophic Models
- Collected Papers, Vol. 5
- Arabic English Glossary
- FlorentinSmarandache
- Data Hiding in Audio Files
- New Techniques to Analyse the Prediction of Fuzzy Models
- Neutrosophic Theory and its Applications, Collected Papers, Vol. 1
- New Research on Neutrosophic Algebraic Structures
- Multi-matrices and arithmetical operations with multi-matrices, by Constantin Scheau
- Complete Synchronization Of Hyperchaotic Xu And Hyperchaotic Lu Systems Via Active Control
- Solving Diophantine Equations
- Progress in Physics, Vol. 1, 2015
- Scientia Magna, Vol. 10, No. 1, 2014
- Neutrosophic Sets and Systems, Vol.-6, -2014
- International Journal of Mathematical Combinatorics, Vol. 4, 2014
- An Introduction to Dynamical Systems and Chaos
- Correspondence Principle
- LaSalle’s Invariance Principle & Chetaev’s Theorem copy
- 07 Witrant Springer
- C0603021013.pdf

Anda di halaman 1dari 100

2014

Ministerul Educaiei

"Aurel Vlaicu" University of Arad, Romania 05~07 November 2014

Research and Education in Innovation Era

5th Edition

Research and Education in Innovation Era

5th Edition

Arad, November 5th 7th, 2014

ISREIE 2014

Mathematics & Computer Science

Romania 5th 7th November 2014

Edited by

Sorin Ndban

Co-editors:

Adrian Palcu, Codrua Stoica, Marius Tomescu

Research and Education in Innovation Era

Fourth Edition

Mathematics & Computer Science

Arad, 5th 7th November 2014

Organized by "Aurel Vlaicu" University of Arad,

Address: B-dul Revoluiei nr. 77, 310130 Arad, Romania, P.O. BOX 2/158 AR

Telephone:+40-257-283010

Fax:+40-257-280070

Email: isreie.uav@gmail.com

http://www.uav.ro/en/conferences/isreie-2014

Copyright 2014 by Editura Universitii Aurel Vlaicu Arad

SCIENTIFIC COMMITTEE

Prof. Michaela Dina Stnescu, Ph.D., Aurel VlaicuUniversity of Arad

Prof. Adrian Nicolescu, Ph.D.,Politehnica University, Bucharest

Assoc. Prof. Alberto Folchini,Ph.D., Politecnico di Milano

Prof. Artur Cavaco-Paulo, Ph.D., Minho University

Prof. Bahrim Gabriela, Ph.D., "Dunarea de Jos"University, Galati

Prof. Burkhard Corves, Ph.D., RWTH Aachen

Doctor Craig Faulds, Food Research Institute, Norwich

Prof. Crisan Popescu, Ph.D., DWI at RWTH, Aachen

Prof. Gabriela Stanciulescu, Ph.D., Bucharest Academy of Economic Studies

Acad. Grozdanka Gojkov, Belgrade University

Prof. Hartwig Hocker, Ph.D.,RWTH Aachen

Prof. Iacob Hantiu, Ph.D., Romanian Council of Sport Science

Prof. Ioan I. Lador, Ph.D., Ministry of Education, Research, Youth and Sport, Bucharest

Prof. Ioan Oprean, Ph.D., University Babes BolyaiUniversity, Cluj-Napoca

Prof. Jannick P. Rolland, Ph.D., University of Rochester, New York

Prof. Laura Muresan, Ph.D., Bucharest Academy of Economic Studies

Prof. Lieva van Langenhove, Ph.D., Ghent University

Prof. Luis Lezaun Martinez de Ubago, Ph.D., Zaragoza University

Prof. Miron Ionescu, Ph.D., Babes-Bolyai University, Cluj-Napoca

Prof. Nicanor Ursua Lezaun, Ph.D.del Pais Vasco/EHU, Bilbao University

Prof. Nicolae Mitrofan, Ph.D., Bucharest University.

Prof. Nicolae Pop, Ph.D. , Bucharest Academy of Economic Studies

Prof. Nicolae Ursu Fischer, Ph.D., Technical University of Cluj-Napoca

Prof. Nicolae Vaszilcsin, Ph.D., Politechnica University, Timisoara

Prof. Pierre de Hillerin, Ph.D., Romanian Institute of Sport Research

Prof. Paul Kickens,Ph.D, Ghent University

Prof. Paulo Brito,Ph.D., Polytechnic Institute, Portalegre

Prof. Renato Rizzo, Ph.D., Naples Federico II University

Prof. Ricardo Ferreira, Ph.D., Polytechnic Institute, Portalegre

Prof. Rodica Zafiu, Ph.D., Bucharest University

Prof. Vincent Nierstrasz, Ph.D., Ghent University

Prof. Viorel Lefter, Ph. D. , Bucharest Academy of Economic Studies

Prof. Werner Reichart, Ph.D., Zurich University

ORGANIZING COMMITTEE

Prof. Florentina-Daniela Munteanu, Ph.D.

Prof. Alina Zamfir, Ph.D.

Prof. Anton Ilica, Ph.D.

Prof. Marian Spnu, Ph.D.

Prof. Dorin Herlo, Ph.D.

Prof. Florea Lucaci, Ph.D.

Prof. Ioan Tulcan, Ph.D.

Assoc. Prof. Pstorel Gapar, Ph.D.

Assoc. Prof. Cecilia Srghie, Ph.D.

Assoc. Prof. Virgil Ciutin, Ph.D.

Assoc. Prof. Dan Glvan, Ph.D.

Assoc. Prof. Alina Roman, Ph.D.

Assoc. Prof. Gabriela Kelemen, Ph.D.

Assoc. Prof. Florin Isac, Ph.D.

Assoc. Prof. Ioan Galea, Ph.D.

Assoc. Prof. Sorin Ndban, Ph.D.

Assoc. Prof. Lcrmioara Ionescu Ph.D

Assist. Prof. Marius Tomescu Ph.D

Reviewers committee

Assoc. Prof. Ph.D. Pstorel Gapar

Assoc. Prof. Ph.D. Adrian Palcu

Assoc. Prof. Ph.D. Codrua Stoica

Contents

trascu Borlea

Codrut

a Stoica, Diana Pa

Stability issues in infinite dimensional spaces . . . . . . . . . . . . . . . . . . . . . . . . .

Luhn prime numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Mihai-Gabriel Babut

ia, Traian Ceausu

Connections between some well-known concepts of uniform exponential

dichotomy for discrete-time systems in Banach spaces . . . . . . . . . . . . . . . . 15

Dominic Bucerzan, Crina Ratiu

Secret communication using cryptography and steganography . . . . . . . . . . 22

Violeta Chis

Fuzzy logic applications in power systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Octavian Cira

Inverse narcissistic numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

36

Carmen T

erei, Marius Tomescu

A study on classification algorithms for imbalanced datasets . . . . . . . . . .

42

Carmen T

erei, Marius Tomescu

Techniques of evaluating classification algorithms for imbalanced data . 48

Mariana Nagy, Negrusa Adina, Burca Valentin

Clusters describing IFRS adoption stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

storel Gaspar, Lorena Popa

Pa

Series representations for random distribution fields . . . . . . . . . . . . . . . . . . 63

Mot

Ghiocel, Popa Lorena

On quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

70

Three concepts of uniform polynomial dichotomy for discrete-time linear

systems in Banach spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

storel Gaspar

Lavinia Sida, Pa

The T -convolution product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

S

andru Andrea, Palcu Adrian

Applications of graph theory in chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . .

91

dimensional spaces

Codruta Stoica

Abstract

The aim of this paper is to define various concepts of stability for cocycles over non-autonomous dynamical systems, as generalizations of the

skew-evolution semiflows, such as exponential stability, (, )exponential

stability, and two more general concepts, the (h, k)stability and the

(h, k)integral stability. Connections between these notions are also given.

Mathematics Subject Classification: 34D05, 93D20

Keywords: Dynamical systems, skew-evolution cocycles, exponential stability,

(, )exponential stability, (h, k)stability, (h, k)integral stability

Preliminaries

In recent years, several concepts of the control theory, as stability, stabilizability, controllability or observability were refined, based on the fact that

the dynamical systems which describe processes from engineering, physics,

biology or economics are extremely complex and the identification of the

proper mathematical models is difficult.

In the qualitative theory of evolution equations, the exponential stability and instability are two of the most important asymptotic properties,

approached lately from various perspectives. Recently, other asymptotic

behaviors were studied, the dichotomy and the trichotomy, by generalizing

the techniques used in the investigation of the stability.

The stability theory puts into discussion the stability of solutions of

differential equations and of trajectories of dynamical systems under small

perturbations of initial conditions. Many parts of the qualitative theory of

differential equations and dynamical systems deal with asymptotic properties of solutions/ trajectories what happens with the system after a long

period of time. If an orbit is well understood, it is natural to ask whether

a small change in the initial condition will lead to similar behavior.

The study of the behaviors of the evolution equations by means of

associated operator families has allowed to obtain answers to open problems by involving techniques of functional analysis and operator theory.

codruta.stoica@uav.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

dianab268@yahoo.com Department of Mathematics, West University of Timi

soara, Romania

Remarkable results of the stability theory are due to J.L. Daleckii and

M.G. Krein (see [1]) and J.L. Massera and J.J. Sch

affer (see [2]).

The concept of evolution operators arises naturally from the theory

of well-posed non-autonomous Cauchy problems, while the notion of linear skew-product semiflows is involved when considering the linearization

along an invariant manifold of a dynamical system generated by a nonlinear differential equation. The skew-evolution semiflows, defined in [3],

are generalizations for both evolution operators and skew-product semiflows. The applicability of this notion was emphasized by P. Viet Hai in

[6] and [7], and by T. Yue, X.Q. Song, D.Q. Li in [8].

We consider in this paper the case of cocycles over multivalued dynamical systems (see [5]). We define various concepts of stability, characterizations for these notions, as well as connections between them.

Definitions. Examples

X = (X, dX ) denotes a complete metric space, P(X) the set of all nonempty subsets of X, V a Banach space with the dual V , B(V ) the space

of all bounded linear operators from V into itself. idX is the identity

map on X, I the identity operator

on V and Y = X V . We consider

R+ such that f (t) = et } and

the sets

E

=

{f

:

R

[1,

)

+

2

T = (t, s) R+ , t s .

Definition 2.1 A multivalued map u

e : T X P(X) with

(s1 ) u

e(t, t, ) = idX , (t, x) R+ X;

(s2 ) u

e(t, t0 , x) u

e(t, s, u

e(s, t0 , x)), (t, s), (s, t0 ) T, x X is a generalized multivalued nonautonomous dynamical system on X.

Remark 2.2 In what follows we will consider the case of a mapping

u : T X X with the properties

(s01 ) u(t, t, x) = x, (t, x) R+ X;

(s02 ) u(t, t0 , x) = u(t, s, u(s, t0 , x)), (t, s), (s, t0 ) T, x X, called

the semiflow associated to the generalized multivalued nonautonomous

dynamical system on X.

Definition 2.3 A mapping U : T X B(V ) which satisfies

(c1 ) U (t, t, x) = I, (t, x) R+ X;

(c2 ) U (t, s, u(s, t0 , x))U (s, t0 , x) = U (t, t0 , x), (t, s), (s, t0 ) T, x X,

is a skew-evolution cocycle over u.

Example 2.4 If U is a skew-evolution cocycle over u and R, then

U : T X B(V ), U (t, t0 , x) = e(tt0 ) U (t, t0 , x) is the shifted

skew-evolution cocycle over u.

Definition 2.5 A skew-evolution cocycle U has growth if there exists

: R+ [1, ), a nondecreasing function, with lim (t) = , that

t

satisfies kU (t, t0 , x)vk (t s) kU (s, t0 , x)vk, for all (t, s), (s, t0 ) T ,

(x, v) Y.

Remark 2.6 (i) If a skew-evolution cocycle U has growth, then U ,

> 0, has also growth.

(ii) The growth is equivalent with the exponential growth (see [4]).

every (t, t0 , x, v ) T X V the mapping s 7 kU (t, s, u(s, t0 , x)) v k

is measurable on [t0 , t].

3

3.1

Stability types

Exponential stability

exist a mapping N : R+ [1, ) and a constant > 0 such that the relation e(ts) kU (t, t0 , x)vk N (s) kU (s, t0 , x)vk holds for all (t, s), (s, t0 )

T , (x, v) Y.

Proposition 3.2 A skew-evolution cocycle U has growth if and only if

there exists > 0 such that the -shifted skew-evolution cocycle U is

exponentially stable.

Proof. Necessity. As U has growth, according to Remark 2.6, there exist M 1 and > 0 such that e(ts) kU (s, t0 , x)vk M kU (t, t0 , x)vk,

for all (t, s), (s, t0 ) T , (x, v) Y . If we consider = 2 > 0 we obtain

kU (t, t0 , x)vk = e(tt0 ) kU (t, t0 , x)vk

M e(tt0 ) e(ts) kU (s, t0 , x)vk =

= M e(tt0 ) e(ts) kU (s, t0 , x)vk e(st0 ) =

= M e(ts) kU (s, t0 , x)vk ,

for all (t, s), (s, t0 ) T and (x, v) Y , which shows that U is exponentially stable.

Sufficiency. As U is exponentially stable, there exist N : R+

[1, ) and > 0 such that e(ts) kU (t, t0 , x)vk N (s) kU (s, t0 , x)vk, for

all (t, s), (s, t0 ) T , (x, v) Y. Further, we obtain

kU (t, t0 , x)vk = e(tt0 ) kU (t, t0 , x)vk

N (s)e(tt0 ) e(ts) kU (s, t0 , x)vk = N (s)e(ts) e(ts) kU (s, t0 , x)vk ,

for all (t, s), (s, t0 ) T and (x, v) Y . We denote

, if >

=

1,

if ,

and we define : R+ [1, ) by ( ) = N ( )e . Hence, U has

growth.

3.2

(, )exponential stability

if there exist some constants N 1, , > 0 such that following relation

e(ts) kU (t, t0 , x)vk N es kU (s, t0 , x)vk holds for all (t, s), (s, t0 ) T ,

(x, v) Y.

Proposition 3.4 A skew-evolution cocycle U with growth is exponentially stable if and only if there exists a constant > 0 such that the

-shifted skew-evolution cocycle U is (, )exponentially stable.

Proof. It is analogous to the proof of Proposition 3.2.

3.3

there exist N 1 and two continuous mappings h, k : R+ R+ such

that

h(t s) kU (t, t0 , x)vk N k(s) kU (s, t0 , x)vk ,

for all (t, s), (s, t0 ) T , (x, v) Y.

Definition 3.6 A skew-evolution cocycle U is (h, k)integrally stable if

there exist a constant D 1 and two continuous mappings h, k : R+

R+ , where h is nondecreasing with the property h(s + t) h(s)h(t), for

all s, t R+ , such that the relation

Z t

h( s) kU (, t0 , x)vk d Dk(s) kU (s, t0 , x)vk

s

3.4

Connections

exponential stability

Remark 3.7 (, )exponentially stable = exponentially stable ([4]).

6=

then U is exponentially stable;

2) If a skew-evolution cocycle U is (h, k)stable and h, k E are given

by t 7 et respectively t 7 M et , M 1 and > , then U is (, )

exponentially stable.

3.4.2

growth is exponentially stable if and only if there exists a constant > 0

such that the -shifted skew-evolution cocycle U is integrally stable.

Proof. Necessity. The existence of a mapping N : R+ R+ and of

a constant > 0 such that kU (t, t0 , x)vk N (s)e(ts) kU (s, t0 , x)vk,

for all (t, s), (s, t0 ) T and (x, v) Y , is assured by Definition 3.1. We

2

Z

Z

e 2 (st0 ) kvk ds =

t0

t0

Z

= N (t0 )

0

e (t0 ) kvk ,

e 2 kvk d N

for all (t0 , x, v) R+ Y , where we have denoted N

Sufficiency. The -shifted skew-evolution cocycle U with > 0 is

integrally stable, with growth, hence, it is stable. Then there exists a

f : R+ R+ such that e(tt0 ) kU (t, t0 , x)vk M

f(t0 ) kvk, for

mapping M

all (t, t0 ) T and (x, v) Y , which proves the exponential stability of

the skew-evolution cocycle U and ends the proof.

Theorem 3.10 A (h, k)integrally stable skew-evolution cocycle U with

growth is (h, k)stable.

Proof. According to Definition 2.5 and Remark 2.6, there exist some constants M 1 and > 0 such that kU (t, t0 , x)vk M e(ts) kU (s, t0 , x)vk,

for all (t, s), (s, t0 ) T , (x, v) Y .

Let t [s, s+1). We have h(ts) kU (t, t0 , x)vk M e h(1) kU (s, t0 , x)vk,

for all (x, v) Y.

Let now s [t 1, t]. We obtain successively

Z t

h(t s) kU (t, t0 , x)vk d

h(t s) kU (t, t0 , x)vk =

t1

t

t1

M e h(1)

s

Conclusions

biology, economics, physics, chemistry and many other fields, deal with

the analysis of the long-time behavior. One of the central interests in

the asymptotic behavior of dynamical systems is to find conditions for

their solutions to be stable, unstable or exponential dichotomic. The

skew-evolution cocyles generalize the notions of semigroups of operators,

evolution families, evolution operators or skew-product semiflows, being

suitable to approach the study of the asymptotic properties of the evolution equations from a non-uniform point of view.

The generality of the concepts presented in this paper is given by the

fact that in the definitions of the growth, of the (h, k)stability and

(h, k)integral stability, exponentials are not necessarily involved.

Acknowledgements. The authors gratefully acknowledge helpful

suggestions from Professor Mihail Megan. We would also like to express

our special thanks to the referees for their support in preparing this paper.

References

[1] J.L. Daleckii, M.G. Krein, Stability of Solutions of Differential Equations in Banach Spaces, Translations of Mathematical Monographs

43, Amer. Math. Soc., 1974.

[2] J.L. Massera, J.J. Sch

affer, Linear Differential Equations and Function Spaces, Pure Appl. Math. 21, Academic Press, 1966.

[3] M. Megan, C. Stoica, Exponential instability of skew-evolution semiflows in Banach spaces, Studia Univ. Babes-Bolyai Math. LIII(1)

(2008), 1724.

[4] C. Stoica, Uniform Asymptotic Behaviors for Skew-evolution Semiflows on Banach Spaces, Mirton Publishing House, Timisoara, 2010.

[5] J. Valero, About non-autonomous multivalued dynamical systems

and their applications, Joint SIAM/RSME-SCM Meeting on Emerging Topics in Dynamical Systems and Partial Differential Equations,

DSPDEs10, Barcelona, 2010.

[6] P. Viet Hai, Continuous and discrete characterizations for the uniform exponential stability of linear skew-evolution semiflows, Nonlinear Anal. 72(12) (2010), 43904396.

[7] P. Viet Hai, Discrete and continuous versions of Barbashin-type theorem of linear skew-evolution semiflows, Appl. Anal. 90(11-12) (2011),

18971907.

[8] T. Yue, X.Q. Song, D.Q. Li, On weak exponential expansiveness of

skew-evolution semiflows in Banach spaces, J. Inequal. Appl. DOI:

10.1186/1029-242X-2014-165 (2014), 16.

Octavian Cira

Florentin Smarandache

Abstract

The first prime number with the special property that its addition with

reversal gives as result a prime number too is 299. The prime numbers

with this property will be called Luhn prime numbers. In this article

we intend to present a performing algorithm for determining the Luhn

prime numbers. Using the presented algorithm all the 50598 Luhn prime

numbers have been, for p prime smaller than 2 107 .

Introduction

The number 299 is the smallest prime number that added with his reverse gives

as result a prime number, too. As 1151 = 229 + 922 is prime.

The first that noted this special property the number 229 has, was Norman

Luhn (after 9 February 1999), on the Prime Curios website [15]. The prime

numbers with this property will be later called Luhn prime numbers.

In the Whats Special About This Number? list [10], a list that contains all

the numbers between 1 and 9999; beside the number 229 is mentioned that his

most important property is that, adding with reversal the resulting number is

prime too.

The On-Line Encyclopedia of Integer Sequences, [9, A061783], presents a list

1000 Luhn prime numbers. We owe this list to Harry J. Smith, since 28 July

2009. On the same website it is mentioned that Harvey P. Dale published on 27

November 2010 a list that contains 3000 Luhn prime numbers and Bruno Berselli

published on 5 August 2013 a list that contains 2400 Luhn prime numbers.

Smarandaches function

with the property that n divides m! (or m! is a multiple of n) is know in the

specialty literature as Smarandaches function, [5, 6, 12]. The values resulting

from n = 1, 2, . . . , 18 are: 1, 2, 3, 4, 5, 3, 7, 4, 6, 5, 11, 4, 13, 7, 5, 6, 17, 6. These

octavian.cira@uav.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Rom

ania

fsmarandache@gmail.com Mathematics & Science Department, University of New Mexico,

USA

values were obtained with an algorithm that results from s definition. The

program using this algorithm cannot be used for n 19 because the numbers

19!, 20!, . . . are numbers which exceed the 17 decimal digits limit and the classic

computing model (without the arbitrary precisions arithmetic [13]) will generate

errors due to the way numbers are represented in the computers memory.

Kempners algorithm

n = pp11 pp22 pps s , prime number and the generalized numeration base (i )[pi ] ,

for i = 1, s, [1]. Partial solutions to the algorithm for (n)s calculation have

been given earlier by Lucas and Neuberg, [12].

Remark 3.1 If n N , n can be decomposed in a product of prime numbers

2

s

1

n = p

1 p2 ps , were pi are prime numbers so that p1 < p2 < . . . < ps , and

s 1, thus Kempners algorithm for calculating the function is.

(n) = max p1 1[p1 ]

(p1 )

, p2 2[p2 ]

(p2 )

, . . . , ps s[ps ]

(ps )

where by [p] (p) we understand that is written in the numeration base p

(noted [p] ) and it is read in the p numeration base (noted (p) , were =

[p] ), [6, p. 39].

Programs

The list of prime numbers was generated by a program that uses the Sieve of

Eratosthenes the linear version of Pritchard, [3], which is the fastest algorithm

to generate prime numbers until the limit of L, if L 108 . The list of prime

numbers until to 2 107 is generated in about 5 seconds. For the limit L > 108

the fastest algorithm for generating the prime numbers is the Sieve of Atkin,

[7].

Program 4.1 The Program for the Sieve of Eratosthenes, the linear version of

Pritchard using minimal memory space is:

CEP bm(L) := f loor L2

f or k 1..

is primek 1

prime (2 3 5 7)T

i last(prime) + 1

f or j 4, 7..

is primej 0

k3

s (primek1 )2

t (primek )2

while t L

f or j t, t + 2 primek ..L

is prime j1 0

2

f or j s + 2, s + 4..t 2

if is prime j1 = 1

2

primei j

ii+1

st

k k+1

t (primek )2

f or j s + 2, s + 4..L

if is prime j1 =1

2

primei j

ii+1

return prime

Program 4.2 The factorization program of a natural number; this program

uses the vector p representing prime numbers, generated with the Sieve of Eratosthenes. The Sieve of Eratosthenes is called upon trough the following sequence:

L := 2 107

t0 = time(0)

(t1 t0 )s = 5.064s

p := CEP bm(L)

last(p) = 1270607

t1 = time(1)

plast(p) = 19999999

j1

k0

f (1 1)

while m pj

if mod (m, pj )=0

k k+1

m

m

pj

otherwise

f stack[f, (pj , k)] if k > 0

j j+1

k0

f stack[f, (pj , k)] if k > 0

return submatrix(f, 2, rows(f ), 1, 2)

We presented the Kempners algorithm using Mathcad programs required

for the algorithm.

Program 4.3 The function counting all the digits in the p base of numeration

in which is n.

10

return 1 otherwise

Where ceil(x) is a Mathcad function which gives the smallest integer x

and log(n, p) is logarithm in base p from n.

Program 4.4 The program intended to generate the p generalized base of numeration noted by Smarandache [p] for a number with m digits.

a(p, m) := f or i 1..m

pi 1

ai

p1

return a

Program 4.5 The program intended to generate for the p base of numeration

noted by Smarandache (p) to write the number.

b(, p) := return (1) if p = 1

f or i 1..ncb(, p)

bi pi1

return b

Program 4.6 Program that determines the digits of the generalized base of

numeration [p] for the number n.

N bg(n, p) := m ncb(n, p)

a a(p, m)

return (1) if m=0

f or i m..1

n

ci trunc

ai

n mod (n, ai )

return c

Where trunc(x) returns the integer part of x by removing the fractional part,

and mod(x, y) returns the remainder on dividing x by y (x modulo y).

Program 4.7 Program for Smarandaches function.

(n) := return Err. n nu este intreg if n 6= trunc(n)

return Err. n < 1 if n < 1

return (1) if n=1

f F a(n)

p f h1i

f h2i

f or k = 1..rows(p)

k pk N bg(k , pk ) b(k , pk )

return max()

11

This program calls the F a(n) factorization with prime numbers. The program

uses the Smarandaches Remark 3.1 about the Kempner algorithm. The .prn

file generation is done once. The reading of this generated file in Mathcads

documents results in a great timesave.

Program 4.8 Program with which the file .prn is generated

V F (N ) := 1 1

f or n 2..N

n (n)

return

This program calls the program 4.7 for calculating the value of the function.

The sequence of the .prn file generation is:

t0 := time(0) W RIT EP RN (.prn) := V F (2 107 ) t1 := time(1)

(t1 t0 )sec = 5 : 17 : 32.625hhmmss

Smarandaches function is important because it characterizes prime numbers

through the following fundamental property:

Theorem 4.9 Let be p an integer > 4, than p is prime number if and only if

(p) = p.

Proof. See [6, p. 31].

Hence, the fixed points of this function are prime numbers (to which is added

4). Due to this property the function is used as primality test.

Program 4.10 Program for testing s primality. This program returns the 0

value if the number is not prime number and the 1 value if the number is a

prime. The file .prn will be read and it will be assigned to the vector.

ORIGIN := 1

:= READP RN ( . . . \.prn)

/ Z if n < 1 n 6= trunc(n)

if n > 4

return 0 if n 6= n

return 1 otherwise

otherwise

return 0 if n=1 n=4

return 1 otherwise

Program 4.11 Program that provides the reveres of the given m number.

R(m) := n f loor(log(m))

x m 10n

f or k 1..n

12

ck trunc(x)

x (x ck ) 10

cn+1 round(x)

Rm 0

f or k n + 1..2

Rm (Rm + ck ) 10

return Rm + c1

Where f loor(x) returns the greatest integer x and round(x) returns x rounded

to the nearest integer.

Program 4.12 Search program for the Luhn prime numbers.

P Luhn(L) := n last(p)

S (229)

k 51

while pk L

N R(pk ) + pk

S stack(S, pk ) if T p(N ) = 1

k k+1

return S

The function stack(A, B, . . .) is applied for merging matrixes top-down. The

number of columns in matrixes should also be the same. The discussed functions

could be applied to vectors as well.

Execution of the program P Luhn was made with sequence

S := P Luhn(2 107 )

The initialization of the S stack is done with the vector that contains the

number 229. The variable k has the initial value of 51 because the index of the

229 prime number is 50, so that the search for the Luhn prime numbers will

begin with p51 = 233.

We present a partial list of the 50598 Luhn prime numbers smaller than 2 107

(the first 321 and the last 120):

229 239 241 257 269 271 277 281 439 443 463 467 479 499 613 641 653 661 673

677 683 691 811 823 839 863 881 20011 20029 20047 20051 20101 20161 20201

20249 20269 20347 20389 20399 20441 20477 20479 20507 20521 20611 20627

20717 20759 20809 20879 20887 20897 20981 21001 21019 21089 21157 21169

21211 21377 21379 21419 21467 21491 21521 21529 21559 21569 21577 21601

21611 21617 21647 21661 21701 21727 21751 21767 21817 21841 21851 21859

21881 21961 21991 22027 22031 22039 22079 22091 22147 22159 22171 22229

22247 22291 22367 22369 22397 22409 22469 22481 22501 22511 22549 22567

13

22571 22637 22651 22669 22699 22717 22739 22741 22807 22859 22871 22877

22961 23017 23021 23029 23081 23087 23099 23131 23189 23197 23279 23357

23369 23417 23447 23459 23497 23509 23539 23549 23557 23561 23627 23689

23747 23761 23831 23857 23879 23899 23971 24007 24019 24071 24077 24091

24121 24151 24179 24181 24229 24359 24379 24407 24419 24439 24481 24499

24517 24547 24551 24631 24799 24821 24847 24851 24889 24979 24989 25031

25057 25097 25111 25117 25121 25169 25171 25189 25219 25261 25339 25349

25367 25409 25439 25469 25471 25537 25541 25621 25639 25741 25799 25801

25819 25841 25847 25931 25939 25951 25969 26021 26107 26111 26119 26161

26189 26209 26249 26251 26339 26357 26417 26459 26479 26489 26591 26627

26681 26701 26717 26731 26801 26849 26921 26959 26981 27011 27059 27061

27077 27109 27179 27239 27241 27271 27277 27281 27329 27407 27409 27431

27449 27457 27479 27481 27509 27581 27617 27691 27779 27791 27809 27817

27827 27901 27919 28001 28019 28027 28031 28051 28111 28229 28307 28309

28319 28409 28439 28447 28571 28597 28607 28661 28697 28711 28751 28759

28807 28817 28879 28901 28909 28921 28949 28961 28979 29009 29017 29021

29027 29101 29129 29131 29137 29167 29191 29221 29251 29327 29389 29411

29429 29437 29501 29587 29629 29671 29741 29759 29819 29867 29989 . . .

8990143 8990209 8990353 8990441 8990563 8990791 8990843 8990881 8990929

8990981 8991163 8991223 8991371 8991379 8991431 8991529 8991553 8991613

8991743 8991989 8992069 8992091 8992121 8992153 8992189 8992199 8992229

8992259 8992283 8992483 8992493 8992549 8992561 8992631 8992861 8992993

8993071 8993249 8993363 8993401 8993419 8993443 8993489 8993563 8993723

8993749 8993773 8993861 8993921 8993951 8994091 8994109 8994121 8994169

8994299 8994463 8994473 8994563 8994613 8994721 8994731 8994859 8994871

8994943 8995003 8995069 8995111 8995451 8995513 8995751 8995841 8995939

8996041 8996131 8996401 8996521 8996543 8996651 8996681 8996759 8996831

8996833 8996843 8996863 8996903 8997059 8997083 8997101 8997463 8997529

8997553 8997671 8997701 8997871 8997889 8997931 8997943 8997979 8998159

8998261 8998333 8998373 8998411 8998643 8998709 8998813 8998919 8999099

8999161 8999183 8999219 8999311 8999323 8999339 8999383 8999651 8999671

8999761 8999899 8999981

Conclusions

The list of all Luhn prime numbers, that totalized 50598 numbers, was determined within a time span of 54 seconds, on an Intel processor of 2.20 GHz.

References

[1] A. J. Kempner, Miscellanea, Amer. Math. Monthly, 25 (1918), 201210.

[2] B. J. Rosser and L. Schoenfeld,Aproximate Formulas for Some Functions of

Prime Numbers Illinois J. Math., 6, (1962), 6494.

14

[3] P. Pritchard, Linear prime number sieves: a family tree, Sci. Comp. Prog.,

9 (1), 1735.

[4] D. Shanks, Solved and Unsolved Problems in Number Theory, Chelsea, New

York, 1993.

[5] F. Smarandache, O nou

a funct, ie n teoria analitic

a a numerelor, An. Univ.

Timis, oara, XVIII fasc.1, (1980), 7988.

[6] F. Smarandache, Asupra unor noi funct, ii n teoria numerelor, Universitatea

de Stat Moldova, 1999, Chis, inau, Republica Moldova.

[7] A. O. L. Atkin and D. J. Bernstein, Prime Sieves Using Binary Quadratic

Forms, Math. Comp., 73, (2004), 10231030.

[8] N. Luhn, pers. comm., 24 Iul. 2014, e-mail:n.luhn@yahoo.de

[9] N. J. A Sloane, Primes p such that p + (p reversed) is also a prime. From:

The On-Line Encyclopedia of Integer Sequences,

http://oeis.org/, 2014 Feb.

[10] E. Friedman, Whats Special About This Number? From: Erichs Place,

http://www2.stetson.edu/efriedma/, 2014 Feb.

[11] E. W. Weisstein, Prime Counting Function. From MathWorldA Wolfram

Web Resource, 2014 Feb.,

http://mathworld.wolfram.com/PrimeCountingFunction.html.

[12] J. Sondow and E. W. Weisstein, Smarandache Function,

http://mathworld.wolfram.com/SmarandacheFunction.html, 2014.

[13] D. Uznanski, Arbitrary precision,

http://mathworld.wolfram.com/ArbitraryPrecision.html, 2014.

[14] Ch. K. Caldwell, The Prime Pages (prime number research, records and

resources),

http://primes.utm.edu/, 2014 Feb.

[15] Ch. K. Caldwell and G. L. Honacher Jr., Prime Curios! The Dictionary of

Prime Number Trivia,

http://www.primecurios.com/, 2014 Feb.

of uniform exponential dichotomy for

discrete-time systems in Banach spaces

Mihai-Gabriel Babutia

Traian Ceausu

Abstract

The aim of this paper is to unify our recent works concerning three

concepts of uniform exponential dichotomy defined for linear discrete-time

systems in Banach spaces: uniform strong exponential dichotomy, uniform

exponential dichotomy,uniform weak exponential dichotomy. By defining

them in the case of strongly invariant sequences of projections, we point

out the connections between them and we also make reference to previous

works in this field, by linking them as well with the concepts defined for

noninvertible systems.

Mathematics Subject Classification: 34D09, 39A05

Keywords: uniform exponentia dichotomy; uniform strong exponential dichotomy;

uniform weak exponential dichotomy; discrete linear system.

Introduction

study of the asymptotic behaviors of discrete-time linear systems in Banach spaces. Several characterizations of such concepts and interesting

results were obtained recently, and we point out a selection of them, as

well as the references within: [1], [2], [3], [4], [6], [7], [9], [11]. A natural

generalization of the dichotomy property, when it cannot completely describe the systems behavior due to the presence of a central manifold, is

the property of uniform exponential trichotomy, and several recent developments have been made ([5], [8] and the references therein).

In this paper we define three concepts of uniform exponential dichotomy for discrete-time linear systems in Banach spaces and establish

the connections between them. Using the previous concepts and results

obtained in [1] and [2], we establish the connections that are present between five concepts of uniform exponential dichotomy, two arising from

mbabutia@math.uvt.ro

ceausu@math.uvt.ro Department of Mathematics, West University of Timi

soara, Romania

15

16

the cited sources. We also pose two remaining open problems in successfully linking all the five presented concepts.

Preliminaries

Let X be a Banach space and B(X) the Banach algebra of all bounded

linear operators on X. The norms on X and on B(X) will be denoted

by k k. The identity operator on X is denoted by I. We will denote by

= {(m, n) N2 : m n}.

We consider the linear difference system

xn+1 = An xn ,

(A)

Definition 2.1 The discrete evolution operator associated to the system (A) is defined, for (m, n) , by:

Am1 . . . An ,

if m > n

An

=

(2.1)

m

I,

if m = n

p

p

Remark 2.2 It is obvious that An

m An = Am , for all (m, n), (n, p)

and every solution of (A) satisfies xm = An

x

m n for all (m, n) .

sequence of projections if Pn2 = Pn for all n N.

If P : N B(X) is a sequence of projections, then the sequence

Q : N B(X) defined by Qn = I Pn is also a sequence of projections,

called the complementary sequence of projections of P .

Definition 2.4 A sequence of projections P : N B(X) is called

invariant for the system (A) if for all n N we have that An Pn =

Pn+1 An ;

strongly invariant for the system (A) if it is invariant for (A)

and for all n N, the restriction An : Ker Pn Ker Pn+1 is an

isomorphism;

bounded if there exist M 1 such that for all n N, kPn k M.

If the sequence of projections P : N B(X) is invariant for the system

(A), then we will say that (A, P ) is a dichotomy pair.

Remark 2.5 If the sequence of projections P is strongly invariant for the

system (A) then

(i) for every n N there is an isomorphism Bn from the Ker Pn+1 to

Ker Pn such that An Bn Qn+1 = Qn and Bn An Qn = Qn ;

n

(ii) for all (m, n) there is an isomorphism Bm

: Ker Pm Ker Pn

n

n

n n

with Am Bm Qm = Qm and Bm Am Qn = Qn for all (m, n) .

N B(X) be a sequence of projections which is strongly invariant for the

linear discrete-time system (A).

We will proceed by presenting the three main concepts of uniform exponential dichotomy with strongly invariant sequences of projections.

Definition 3.1 Let (A, P ) be a dichotomy pair in which P : N B(X) is

strongly invariant for the discrete system (A). We say that the dichotomy

pair (A, P ) is:

uniformly strongly exponentially dichotomic (u.s.e.d) if there

exist N 1 and > 0 such that for all (m, n) ,

(used1 )

(mn)

kAn

m Pn k N e

(used2 )

n

kBm

Qm k N e(mn) .

and > 0 such that for all (m, n, x) X,

(ued1 )

(mn)

kAn

kPn xk

m Pn xk N e

(ued2 )

n

kBm

Qm xk N e(mn) kQm xk.

exist N 1 and > 0 such that for all (m, n) ,

(uwed1 )

(mn)

kAn

kPn k

m Pn k N e

(uwed2 )

n

kBm

f Qm k N e(mn) kQm k.

by the following result.

Proposition 3.2 The following assertions hold.

(a) (A, P ) is u.s.e.d if and only if there exist N 1 and > 0 such that

for all (m, n, x) X the following conditions hold:

(used01 )

(mn)

kAn

kxk

m Pn xk N e

(used02 )

n

kBm

Qm xk N e(mn) kxk.

(b) If the dichotomy pair (A, P ) is u.s.e.d then (A, P ) is u.e.d and P :

N B(X) is bounded.

(c) If the dichotomy pair (A, P ) is u.e.d then (A, P ) is u.w.e.d.

(d) If P : N B(X) is bounded and (A, P ) is u.w.e.d then (A, P ) is

u.s.e.d.

Proof. (a) follows straightforward from the definition of the operator

norm on B(X). To prove (b), let (m, n, x) X and N, given by the

u.s.e.d property of the pair (A, P ). The conclusion follows by considering

the vectors Pn x and Qm x in (used01 ) and (used02 ) respectively. Moreover,

by taking m = n in (used1 ) we obtain that kPn k N .

Assertion (c) follows immediately, by taking the operator norm in (ued1 )

and (ued2 ). In order to prove (d), consider (m, n) and N, given by

17

18

for P : N B(X) . From the following estimations

(mn)

kAn

kPn k 2M N e(mn)

m Pn k N e

n

kBm

Qm k N e(mn) kQm k 2M N e(mn)

exist N, > 0 such that for all (m, n, x) X,

(i)

(ii)

(mn)

kAn

kPn xk

m Pn xk N e

m Qn xk.

n

X. For the necessity, we compute kQn xk = kBm

Qm An

m Qn xk

n

N e(mn) kAn

Q

xk.

The

sufficiency

follows

from

kB

Q

m n

m m xk =

n

n

(mn)

kQn Bm

Qm xk N e(mn) kAn

kQm xk.

m Qn Bm Qm xk = N e

Remark 3.4 The above proposition shows us that in both cases in which

P is invariant and strongly invariant for (A) respectively, the dichotomy

concepts are equivalent. In the next section we will see that in the case of

the strong and weak concepts, this is not the case.

Remark 3.5 If a pair (A, P ) is u.e.d, it does not necessarily follow that

(A, P ) is u.s.e.d. The following example points out this fact.

Example 3.6 On X = R2 endowed with the max-norm, consider, for

every n N, Pn , An : R2 R2 defined, for x = (x1 , x2 ) R2 , by

Pn x = (x1 + ln(n + 1) x2 , 0),

An =

1

Pn + eQn+1 ,

e

projections which is strongly invariant for the system (A). The evolution

operator associated to (A) is given by

nm

An

Pn + emn Qm ,

m = e

(m, n) .

nm

Having in mind that for all (m, n, x) X, kAn

kPn xk

m Pn xk = e

n

mn

mn

and kAm Qn xk = e

kQm xk e

kQn xk it follows that the pair

(A, P ) is u.e.d. But (A, P ) cannot be u.s.e.d because, from the fact that

for all n N, kPn (0, 1)k = ln(n + 1), it follows that kPn k ln(n + 1)

hence P : N B(X) is not bounded.

Remark 3.7 From the above example, in light of (iii) from Proposition

3.2, we deduce that the dichotomy pair (A, P ) from Example 3.6 is u.w.e.d

but fails to be u.s.e.d.

Open problem. At this moment we do not know wether the implication (A, P ) is u.w.e.d (A, P ) is u.s.e.d is generally true or not.

For a more facile reading of the paper, the concepts defined in this paragraph will be denoted with a bar over their abbreviations. Although these

concepts can be defined for dichotomy pairs (A, P ) without the invertibility condition on the kernels of the projections, we will assume, as in the

previous section, that P : N B(X) is strongly invariant for the system (A), in order to establish some connections between the concepts

presented throughout this paper.

Definition 4.1 We say that the dichotomy pair (A, P ) is:

u.s.e.d if there exist N 1 and > 0 such that for all (m, n)

(used1 )

(mn)

kAn

m Pn k N e

(used1 )

1 N e(mn) kAn

m Qn k

u.w.e.d if there exist N 1 and > 0 such that for all (m, n)

(uwed1 )

(mn)

kAn

kPn k

m Pn k N e

(uwed1 )

m Qn k

Remark 4.2 The above defined concepts, in the continuous case, were

studied in [1], and the connections between them can easily be established in a similar manner. Moreover, in the general framework of (h, k)dichotomies for discrete-time linear systems, the u.s.e.d concept is used

in [2]. Taking into account the results from these two papers, we state the

following result.

Proposition 4.3

(c) If (A, P ) is u.e.d then (A, P ) is u.w.e.d.

(d) The concepts of u.s.e.d and u.e.d do not imply each other, existing linear discrete-time systems that satisfy one condition, but not

verifying the other.

(e) If (A, P ) is u.w.e.d it does not result that (A, P ) is u.e.d.

(f ) If (A, P ) is u.w.e.d it does not result that (A, P ) is u.s.e.d.

Proof. Assertion (a) follows from the fact tha for all (m, n) , kQn xk

n

kBm

Qm k kAn

m Qn xk. Assertion (b) follows from the fact that if (A, P )

is u.s.e.d then 1 max{kPn k, kQn k} N , for all n N. Assertion (c)

follows by taking the operator norm in (i) and (ii) in Proposition 3.3.

In order to prove assertion (d), we refer to Examples 13 and 14 from

[2], for the particular case in which hn = en = kn , n N. Finally, in

order to prove (e) we refer to Example 4.4. For (f), Example 3.6 gives a

dichotomy pair (A, P ) which is also u.w.e.d but cannot be u.s.e.d, because

P : N B(X) is not bunded.

19

20

Example 4.4 We will give an example in the general case of noninvertible linear systems. On X = R3 endowed with the max-norm, consider

the system (A) described by the operators An : R3 R3 defined, for all

n N and x = (x1 , x2 , x3 ) R3 , by An x = (e1 x1 , ex2 , 0). The evolution

operator associated to the system (A) is given by

(

x

m=n

n

, (m, n, x) R3 .

Am x =

nm

mn

(e

x1 , e

x2 , 0) m n + 1

Define, for every n N and x R3 , Pn x = (x1 , 0, 0). We have that

(A, P ) is a dichotomy pair. Moreover, for all (m, n) , kAn

m Pn k =

mn

e(mn) kPn k and kAn

kQn k, thus (A, P ) is u.w.e.d. By asm Qn k = e

suming that (A, P ) is s.e.d, in particular, for x = (0, 0, 1), we would obtain

that 0 = N e(mn) kAn

m Qn xk kQn xk = 1 which is a contradiction.

Proposition 4.5 If (A, P ) is u.s.e.d then (A, P ) is u.s.e.d.

Proof. It is similar to the proof of Proposition 23 from [2].

Remark 4.6 The converse of the preceding proposition does not generally

hold, and we refer to Example 24 form [2].

Open problem. At this moment, we do not have an answer to the

implication (A, P ) is u.w.e.d (A, P ) is u.w.e.d.

Remark 4.7 Regarding the open problem from above, we have a partial

answer, which states that if (A, P ) is u.w.e.d then it does not necessarily

imply that (A, P ) is u.w.e.d, as it can be seen from Example 4.8.

Example 4.8 On X = R3 endowed with the max-norm, consider, for

n, N, An , Pn : R3 R3 defined, for every x = (x1 , x2 , x3 ) R3 , by

1

An x = ( x1 , ex2 , e(2n+1) x3 ),

e

Pn x = (x1 , 0, 0).

nm

An

x1 , emn x2 , en

m x = (e

m2

x3 ),

(m, n, x) R3 .

nm

strongly invariant for (A) and kAn

kPn k, kAn

m Pn k = e

m Qn (0, 1, 1)k

mn

emn kQn (0, 1, 1)k, from where kAn

Q

k

e

kQn k hence (A, P ) is

m n

u.w.e.d. Assume, by a contradiction, that (A, P ) is u.w.e.d. Then there

exist N 1 and > 0 such that for all (m, n) , we have that

n

kBm

Qm k N e(mn) kQm k. Taking into account that for all (m, n, x)

2

2

n

Qm xk = max{enm |x2 |, em n |x3 |}, for x0 = (0, 0, 1),

R3 , kBm

2

2

n

we get that for all (m, n) , kBm

Qm x0 k = em n kQm x0 k hence

2

2

n

kBm

Qm k em n . We finally obtain that

em

n2

n

kBm

Qm k N e(mn) kQm k = N e(mn) ,

(m, n) ,

Remark 4.9 Taking into account the connections between the studied

concepts, the results from the cited papers and the open problems stated

in this paper, we can emphasize the connections between the presented

concepts throughout the following diagram.

6

u.s.e.d

6

u.s.e.d

u.e.d

m

u.e.d

u.w.e.d

6

u.w.e.d

u.s.e.d

6

u.s.e.d

References

[1] M.-G. Babutia, T. Ceausu, N. M. Seimeanu, On Uniform Exponential

Dichotomy of Evolution operators, An. Univ. Vest Timis., Ser. Mat.Inform., L, 2 (2012), 3-14.

[2] M.-G. Babutia, M. Megan, I.-L. Popa, On (h, k)-dichotomies for

nonautonomous difference equations in Banach spaces, International

Journal of Differential Equations, Vol. 2013 (2013), Arcile ID 761680,

7 pages.

[3] A. J. G. Bento, C. M. Silva, Generalized nonuniform dichotomies and

local stable manifolds, J. Dyn. Diff. Equat. (2013). 25:1139-1158.

[4] M. Megan, A. L. Sasu, B. Sasu, Discrete admissibility and exponential

dichotomy for evolution families, Discrete Contin. Dynam. Systems

9 (2003), 383-397.

[5] M. Megan, C. Stoica, On uniform exponential trichotomy of evolution

operators in Banach spaces, Integ. Equat. Oper. Th. 60 (4) (2008)

499-506.

[6] I.-L. Popa, M. Megan, T. Ceausu, Exponential dichotomies for linear

difference systems in Banach spaces, Appl. Anal. Discrete Math., 6

(2012), 140-155.

[7] I.-L. Popa, M. Megan, T. Ceausu, Nonuniform exponential dichotomies in terms of Lyapunov functions for noninvertible linear

discrete-time systems, The Scientific World Journal, vol. 2013 (2013),

Article ID 901026.

[8] A. L. Sasu, B. Sasu, Discrete admissibility and exponential trichotomy

of dynamical systems, Discrete Contin. Dyn. Syst. 34 (7) (2014),

2929-2962.

[9] B. Sasu, A. L. Sasu, On the dichotomic behavior of discrete dynamical

systems on the half-line, Discrete Contin. Dyn. Syst. 33 (2013), 30573084.

[10] B. Sasu, A. L. Sasu, Exponential dichotomy and (lp , lq )-admissibility

on the half-line, J. Math. Anal. Appl. 316 (2006), 397-408.

[11] C. Stoica, M. Megan, Asymptotic behaviors for evolution equations in

infinite dimensional space, Proceedings of the International Symposium - Research and Education in an Innovation Era, Section Mathematics, 3rd Edition, Aurel Vlaicu University of Arad Publishing

House (2010) 187-191.

21

steganography

Dominic Bucerzan

Crina Ratiu

Abstract

Cryptography and steganography are two techniques used to provide

data confidentiality. Cryptography scrambles the digital data so the

new content becomes impossible to understand to unauthorized users.

Steganography embeds the digital data in a cover file (usually an image

or audio file). Steganography, is a technique that camouflages a communication to hide its existence and make it seem invisible to unauthorized

users. In this paper we present both methods, but we insist on the advantage provided by combining these two methods in order to improve the

security of communication over Internet by means of commonly available

equipments as PCs, tablets and smart phones.

Mathematics Subject Classification: 94A62

Keywords: Cryptography, Steganography, LSB, Android, SmartSteg

Introduction

and software technology. Even though the progress in this fields has occurred with a higher speed than ever before, digital information security

problems remain present and become an interdisciplinary issue, having to

be constantly optimized, developed and innovated [1].

In figure 1 we present briefly a classification of the security techniques used today to ensure the main characteristics that defines digital

information: integrity, confidentiality, availability, authenticity and norepudiation[8].

In the scientific literature, Simmons, in 1983, proposed the Prisoners

Problem to define the digital information security environment [3]. The

Prisoners Problem describes Alice and Bob who are in jail and wish

to establish an escape plan; they may communicate through the warden,

Willie who shouldnt suspect and uncover their plan. In this situation

the two prisoners must find a way to hide their secret communication

in order not to arouse any suspicion from Willie [3]. So Alice and Bob

dominic@bbcomputer.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

ratiu anina@yahoo.com Babes-Bolyai University of Cluj-Napoca, Romania

22

messages.

These two techniques are widely used today to ensure the security of

the information transferred through available channels for communication, which are vulnerable to interception. Moreover in recent years these

techniques were combined and used together to optimize the security of

digital information [1].

Cryptography

it becomes unreadable to unauthorized users in the process of storing or

transferring it to its intended receiver using todays open channels for

communication.

Cryptography is recognized today to be the main component of the

security policy of every organization and considered a standard to digital

data security [4].

Kirchhoff has stated the principle that stands as a basis for every

cryptographic system: the security of an encryption system should rely

on the secrecy of the encryption/decryption key and not on the secrecy of

the algorithm used [4]. This means that if an unauthorized user knows the

algorithm used for encryption, without knowing the key used, is unable to

decrypt the secret data. Having this into consideration we may say that

the security level of a cryptographic system stands in the length of the

key used. The larger the key, the longer is the time needed to uncover it

[4].

According to the type of key used, cryptographic systems are classified

as [5]:

Secret / Symmetric key cryptography: the same key is used for both

processes encryption and decryption. The sender uses the key to

encode the secret data; the receiver uses the same key to decode the

data received.

Public / Asymmetric key cryptography: it uses two key for encrypting and decrypting the data. The sender uses the public key of the

receiver (the key is public, known by anyone) to encrypt the data;

23

the receiver uses his personal secret key (only the receiver knows the

secret key) to decode the received data.

The cryptographic methods mentioned are the basis for two popular

techniques used today in information security technology, namely:

Digital signature: ensures the authentication of the sender

Hash function: ensures the authentication and the integrity of the

sent data.

Steganography

files (usually media files) in such a way that its presence becomes difficult

to notice [1].

Some countries raised restrictions upon the usage of cryptography; in

this scenario steganography became an alternative for confidential communications. Steganography was seen as a solution to this situation because

a steganographic message couldnt be detected so its usage couldnt be

controlled [1].

By comparison with cryptography which only scrambles the message in

a way unreadable for unauthorized users, steganography embeds the data

using a cover file in such a way that the changes made to the cover file are

not obvious to human means of observation and it can pass through open

channels of communication without raising any attention or suspicion.

Steganography is not something new, it is a technique used since antiquity. Nowadays it has evolved and developed to be used with digital

technology. Most used cover files for steganography are media files like

video, audio and especially image files[7].

Image files are proper to steganography due to their structure, to the

fact that the changes made to an image file are difficult to notice with

the naked eye. Also, the technology has evolved and it became cheaper,

hence digital images are omnipresent in everyday life.

One of the most used methods in digital image steganography is the

LSB (Last Significant Bit) technique also known as noise insertion. This

technique usually uses as cover files digital images of RGB type [8]. By

using the last significant bits of the bytes of the original cover file it embeds

the bits of secret information. Here you have an example: consider three

pixels of an 24 bits RGB image using nine bytes of memory.

(001001111110100111001000)

(001001111100100011101001)

(001001111100100011101001)

To embed the character A which has the following bit structure 100000001,

the original bits schem will change as follows:

(001001111110100 [0] 11001000)

(0010011 [0] 110010001110100 [0])

(110010000010011111101001)

24

As it can be seen only the value of three bits needs to change in order

to embed the desired character. These changes are not visible to the

naked eye. In average only a half of the LSB of a digital image needs to

be changed to embed secret data.

steganography

A method to optimize the secrecy of digital information communication

in todays insecure environments like Internet and Mobile Networks is

to combine cryptography with steganography. By hiding the previously

encoded message, the information that is intended to be secret may be

transmitted without attracting attention. Also, this method, offers an

alternative to the classical storage of data. A user may store secret data

by encoding and embedding them in digital images to protect them from

a possible intruder [2].

Both, steganography and cryptography are intended as means to protect the confidentiality of information. However, on their own, they are

not perfect and can be decrypted and revealed. This is why most experts

would suggest the use of multiple layers of security.

The literature does not recommend replacing steganography with cryptography or cryptography with steganography. It is recommended the simultaneous use of the two techniques. This recommendation derives from

the differences between the two techniques and the fact that they can

complete each other. Figure 2, presents briefly the different purposes that

cryptography and steganography fulfill [5].

Current trends, both in hardware and software technology led industry

towards the development of smaller, faster and high-performance mobile

devices, which can support a wide range of features and open source operating systems [2].

A rapid growth in this area is registered by mobile hand-held devices

which are popularly called smart gadgets and they include: smart phones,

tablets, e-book readers. The Smartphones life-cycle has evolved drastically in recent years, having a lifetime of approximately 6 months between

generations [2].

25

Among the major issues that occur in this dynamic, ever changing and

evolving environment is the fact that almost all platforms have dedicated

application. This fact is in contradiction with one of the principal characteristic of digital information, namely availability. In other words, if a

sender uses an iPhone to encode and hide some information, the receiver

must also use an iPhone to unhide and decode the secret information.

And this issue also applies when using a computer [2].

In the case when the sender is restricted to using a computer whereas

the receiver has access only to a smart-phone, the secret information is

no longer available in this scenario.

A proposed solution to this situation is SmartSteg project that includes

a set of applications that use the same algorithm to encode - decode, hide reveal secret information (almost any kind of digital file) in digital images

independent of the type of device used. The idea is shown in figure 3.

At this stage of the project, SmartSteg provides confidential communication between computers that run under Windows Operating System

and devices that run under Android. This means that a secret file that is

encrypted and encoded with SmartSteg using a computer can be revealed

and decrypted with SmartSteg from a device that runs under Android.

Figure 4 and 5 show the design of the application chosen for the two

versions of SmartSteg [1].

Conclusions

provide secret communications. However, they differ in the methods of

attacking / breaking of the systems [6].

A cryptographic system is considered broken if an attacker manages

to decode the secret information. A steganography system is considered

broken if an attacker can detect the existence of the hidden information

26

Figure 4: Design for the application that runs under Windows [1]

Figure 5: Design for the SmartSteg application running under Android [1]

considered to have failed even if an attacker only suspects the existence

of a file with hidden information or the method used to conceal the secret

information even without being able to extract it.

Steganography adds an extra layer of security to cryptography, so combining the two techniques may achieve an efficient solution in terms of

systems that ensure confidential communications [6].

References

[1] D. Bucerzan, C. Ratiu, Image Processing with Android Steganography, Springer Advances in Intelligent Systems and Computing ISSN:

2194-5357

[2] D. Bucerzan, C. Ratiu, Steganography and cryptography on mobile

platforms, Constanta Maritime University Annals, Year XIV, Vol.20,

27

2013.

[3] Ross J. Anderson, Fabien A.P. Petitcolas, On The Limits of

Steganography , IEEE Journal of Selected Areas in Communications,

16(4):474-481, May 1998. Special Issue on Copyright & Privacy Protection. ISSN 0733-8716

[4] B B Zaidan, A.A Zaidan, A.K. Al-Frajat and H.A. Jalab, On the Differences between Hiding Information and Cryptography Techniques:

An Overview, Journal of Applied Sciences 10(15): 1650-1655, 2010

[5] A.Joseph Raphael , V. Sundaram, Cryptography and Steganography A Survey Int. J. Comp. Tech. Appl., Vol 2 (3), 626-630 ISSN:22296093

[6] A. Almohammad, Steganography-Based Secret and Reliable Communications: Improving Steganographic Capacity and Imperceptibility

A thesis submitted for the degree of Doctor of Philosophy, Department

of Information Systems and Computing, Brunel University, August,

2010

[7] Kallam Ravindra Babu, S.Udaya Kumar,A.Vinaya Babu, A Survey

on Cryptography and Steganography Methods for Information Security International Journal of Computer Applications,Volume 12 No.2,

November 2010

[8] Cole E, Hiding in Plain Sight. Steganography and the Art of Covert

Communication (2003)(335), Wiley Publishing, Inc, ISBN: 0-47144449-9

[9] D. Bucerzan, C. Ratiu, SmartSteg: A New Android Based

Steganography Application, International Journal of Computers

Communications & Control, Vol 8, No.5 (2013), available at:

http://univagora.ro/jour/ index.php/ijccc/article/view/642

28

Violeta Chis

Deacu Antheia

Abstract

Fundamentals of fuzzy logic and an overview of its applications in

power systems are presented first. Fuzzy logic based systems with their

capability to deal with incomplete information, imprecision, and incorporation of qualitative knowledge have shown great potential for application

in electric load forecasting. The paper investigates the application of fuzzy

logic (FL) as forecasting tools for predicting short term load forecasting

(STLF).

Mathematics Subject Classification: 34D09

Keywords: Fuzzy logic, short term load forecasting (STLF), power systems.

Introduction

Electric power systems are large, complex, geographically widely distributed systems and influenced by unexpected events. These facts make

it difficult to effectively deal with many power system problems through

strict mathematical approaches. Therefore, intelligent techniques such as

expert systems, artificial neural networks (ANN), genetic algorithms (GA)

and FL have emerged in recent years in power systems as a complement to

mathematical approaches and have proved to be effective when properly

coupled. As the real world power system problems may neither fit the assumptions of a single technique nor be effectively solved by the strengths

and capabilities of a single technique, it is now becoming apparent that

the integration of various intelligent techniques is a very important way

forward in the next generation of intelligent systems [1].

Uncertainty and imprecision widely exist in engineering problems. The

complexity in the world generally arises from uncertainty in the form of

ambiguity. The following lists just some examples of such uncertainty and

imprecision in power systems:

Changing power system operation conditions, such as changes in

load or generation, and changes in the topology of power systems.

Various power system constructions, such as untransposed/transposed,

shunt compensation, and series compensation, especially the introduction of new techniques - flexible AC transmission systems.

viochis@uav.ro Department of Mathematics and Computer Science, Aurel Vlaicu University of Arad, Romania

antheia.deacu@upt.ro Universitatea Politehnica din Timisoara

29

Many different fault/disturbance conditions, including fault inception, fault location, fault types, and fault path resistance.

Inaccuracies caused by voltage and current transducers or SCADA

measurements/state estimations or noise introduced through electromagnetic interference.

Uncertainty caused by electricity market or deregulation.

Imprecise information caused by human beings involved in the planning, management, operation and control of power systems.

Many concepts in power system problems are fuzzy in nature and

the knowledge and reasoning used by human experts to solve these

problems is approximate.

Analytical approaches have been used over the years for many power system operation, planning and control problems. However, the mathematical formulations of real-world problems are derived under certain restrictive assumptions and even with these assumptions, the solutions of large

scale power systems problems are not trivial. On the other hand, there

are many uncertainties in various power systems problems because power

systems are large, complex, geographically widely distributed systems and

influenced by unexpected events. More recently, the deregulation of power

utilities has introduced new issues into the existing problems. These facts

make it difficult to effectively deal with many power systems problems

through strict mathematical formulations alone. Although a large number of AI techniques have been employed in power systems, fuzzy logic is

the only possible answer to a number of challenging problems [1].

For the most complex system where few numerical data exist and only

ambiguous or imprecise information may be available, fuzzy reasoning provides a way to understand system behavior by allowing us to interpolate

approximately between observed input and output situation. In recent

years, the number of publications in the area of fuzzy logic applications

to power systems has been growing rapidly [1]. The areas include:

modelling and control e.g. power system stability control;

pattern recognition and predication, e.g. power system security assessment, fault diagnosis, load forecasting and power system protection;

optimisation, e.g. power system planning, unit commitment and

economic dispatch etc.

economical and reliable operations for power stations and their generating

30

is essential part of any energy management system [2].

The load forecasting can be divided into three categories: long term,

medium term and short term. Long term load forecasting (LTLF) is applicable for system and long term network planning. Mid Term Load

Forecasting (MTLF) refers to quarterly, half yearly and yearly load forecasting needs. Short-term load forecasting (STLF) is used to predict load

demands up to a week ahead so that the day-to-day operation of a power

system can be efficiently planned [3].

There are many techniques that could be employed for loads forecasting like statistical method, linear regression, FL, ANN, GA, expert

system, support vector machine, and data mining model.

Electricity load demand is influenced by many factors, such as weather,

economic and social activities, and different load. By analysis of only historical load data, it is difficult to obtain accurate load demand for forecasting. The relation between load demand and the independent variables

is complex and it is not always possible to fit the load curve using statistical models. The numerical aspects and uncertainties of this problem

appear suitable for fuzzy methodologies [4].

Case Study

of fuzzy model was implemented using Matlab Fuzzy Toolbox. There are

four basic elements in a fuzzy system (Figure 1) which are:

1. Fuzzification: the process of associating crisp input values with the

linguistic terms of corresponding input linguistic variables

2. Fuzzy inference engine: provides the decision making logic of the

system The fuzzy inference system is a popular computing framework based on the concept of fuzzy set theory, fuzzy if-then rules,

and fuzzy reasoning.

3. Fuzzy rule base: a set of linguistic rules or conditional statements in

the form of

IF a set of conditions IS satisfied, THEN a set of consequences are

inferred

4. Defuzzification interface: Defuzzifies the fuzzy outputs of the fuzzy

inference machine and generates a non-fuzzy (crisp) output which

is the actual output of the fuzzy system [5]. The Center of Gravity

method (COG) is the most popular defuzzification technique and is

widely utilized in actual applications.

Our model forecasts the load for one whole day at a time. Fuzzy

logic based model is developed and presented for STLF using real data

(real load data of Romanian electric utility). Weather related variation is

certainly critical in predicting electricity demand for lead times beyond a

day-ahead.

In this case, the inputs are two previous temperature. Temperature

is important because demand of load is depending on temperature of the

day. Normally when temperature is high, the demand will also high.

31

There are two most used types of Fuzzy Inference System (FIS): Mamdanis and Sugenos. These two types of inference systems vary somewhat

in the way the outputs are determined. Mamdanis fuzzy inference method

is the most commonly seen fuzzy methodology due to its simple structure

of min-max operations. Figure 2 shows the whole structure of fuzzy logic

system included input, reasoning rules and also the proposed output.

Construction of membership functions can be based on intuition, experience or probabilistic methods. The two inputs taken for STLF are

t1 and t2. As shown in figure (Figure 3) t1 and t2 is divided into five

triangular membership functions.

Figure 4 shows forecasted load (output) divided into five triangular

membership functions. Fuzzy Rule Base is the heart of the fuzzy system.

The heuristic knowledge of the forecasted is stored in terms of IF-THEN

rules. It sends information to fuzzy inference system, which evaluates

32

the gained information to get the load forecasted output. For the STLF

problem, a set of multiple-antecedent fuzzy rules have been established.

Some of the rules are prezented in Figure 5 as follows:

The Surface Viewer (Figure 6) is used to display the dependency of

one of the outputs on any one or two of the inputs - that is, it generates

and plots an output surface map for the system.

The Rule Viewer (Figure 7) is based on the fuzzy inference diagram

and displays a roadmap of the whole fuzzy inference process.

Results of fuzzy logic based models are compared with the actual demand of electricity for validation. To evaluate the result of fuzzy systems

33

34

error =

|actualload f orecastedload|

100

actualload

(4.1)

The predicted data of electrical load for one day is compared with

actual load demand and presented graphically in Figure 8.

The performance of the model is evaluated on the basis of statistical

indicator; the average error in the forecasted load in comparison with the

desired load is 2.59

Conclusions

systems, fuzzy logic is the only possible answer to a number of challenging

problems.

approaches to power systems clearly demonstrates that fuzzy logic has

been, is, and can be used to solve power systems problems.

The fuzzy logic model for the short term electrical load forecasting is

developed and presented. The developed model is accurate and effective

for short term load forecasting.

References

[1] Momoh J.A., MA., x., and TOMSOVIC, K, Overview and literature

survey of fuzzy set theory in power systems), IEEE Trans. Power Systems,1995, PWRS-10, (3, pp. 1676-1690

[2] Chaturvedi, D.K., Satsangi, P.S. Kalra, P.K, Fuzzified Neural Network

Approach for Load Forecasting Problems, Int. J. on Engineering Intelligent Systems, CRL Publishing, U.K., Vol. 9, No. 1, March 2001,

pp.3-9.

[3] M. Rizwan, Dinesh Kumar, Rajesh Kumar, Fuzzy Logic Approach for

Short Term Electrical Load Forecasting, Electrical and Power Engineering Frontier, 2012, Vol. 1 Iss. 1, pp. 8-12

[4] A. G. Bakirtzis, V. Petridis, S. J. Kiartzis, M. C. Alexiadis and A. H.

Maissis, A neural network short term load forecasting model for the

Greek power system,Power Systems, IEEE Transactions on, vol. 11,

pp. 858-863, 1996.

[5] J. W. Taylor, L.M. de Menezes and P. E. McSharry, A Comparison of

Univariate Methods for Forecasting Electricity Demand up to a Day

Ahead,International Journal of Forecasting, pp 1-16, 2006.

35

Octavian Cira

Abstract

An n-digit number that is the sum of the nth powers of its digits is

called an n-narcissistic number. The sum of n numbers n to powers dk ,

where dk {0, 1, . . . , b 1}, n, b N, 1 < n b are the digits in base b, is

a inverse narcissistic number. In this article determine inverse narcissistic

numbers, in bases of numeration b = 2, 3, . . . , 16.

Introduction

An n-digit number that is the sum of the nth powers of its digits is called an

n-narcissistic number. It is also sometimes known as an Armstrong number,

perfect digital invariant [3], or plus perfect number. Hardy [7] wrote, There

are just four numbers, after unity, which are the sums of the cubes of their

digits: 13 + 53 + 33 = 153, 33 + 73 + 03 = 370, 33 + 73 + 13 = 371, and

43 + 03 + 73 = 407. These are odd facts, very suitable for puzzle columns and

likely to amuse amateurs, but there is nothing in them which appeals to the

mathematician. Narcissistic numbers therefore generalize these unappealing

numbers to other powers [3, page 164]. In the article [14] was determined

Narcissistic numbers in bases of numeration b = 2, b = 3,. . . b = 16. These

numbers are solutions Diophantine equations of order n

dnn1 + dnn2 + . . . + dn0 = dn1 bn1 + dn2 bn2 + . . . + d0 ,

where d0 , d1 , . . . , dn1 {0, 1, . . . , b 1}, n, b N, 1 < n b. These numbers

narcissistic dealt many authors, among which [1, 2, 4, 5, 6, 8, 9, 10, 11, 12, 13,

18, 15, 16, 17].

The solution the equation Diophantine of the order b 1,

ndn1 + ndn2 + . . . + nd0 = dn1 bn1 + dn2 bn2 + . . . + d0

(1.1)

octavian.cira@uav.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Rom

ania

36

37

We use the notation dn1 dn2 . . . d0(b) for number written in the base b, dn1

bn1 + dn2 bn2 + . . . + d0 b0 . For numbers written in base b = 10 we use

classical notation.

The largest number of the form ndn1 + ndn2 + . . . + nd0 in base b is n nb1 .

The smallest number of n digits in base b is bn1 + 1. Let the function h :

{2, 3, . . . , 16} {2, 3, . . .} R

h(b, n) = ln n nb1 ln bn1 + 1 .

(2.1)

If we have h(b, n) > 0 then it follows that Diophantine equation (1.1) may

have solutions, if h(b, n) 0 then equation (1.1) has not solutions. To see the

discrete values (b, n) for which equation (1.1) may have solutions refer to figure

1.

The solutions with n = 2 digits and for b = 16, b = 15, . . . b = 2:

21 + 2 4

14(14) = 18 ,

62(11) = 68 ,

24(8) = 20 ,

13(7) = 10 ,

44(7) = 32 ,

12(4) = 6 ,

10(3) = 3 ,

11(3) = 4 ,

22(3) = 9 ,

2 +2

2 +2

2 +2

2 +2

2 +2

2 +2

2 +2

2 +2

0.25s, where they analyzed a total of 1360 cases.

The solutions with n = 3 digits and for b = 16, b = 15, . . . b = 2:

31 + 3 0 + 3 4

104(9) = 85 ,

500(7) = 245 ,

522(7) = 261 ,

133(6) = 57 ,

343(6) = 135 ,

113(5) = 33 ,

432(5) = 117 ,

213(4) = 39 ,

3 +3 +3

3 +3 +3

3 +3 +3

3 +3 +3

3 +3 +3

3 +3 +3

3 +3 +3

1.684s, where they analyzed a total of 17000 cases.

38

41 + 40 + 40 + 46

1006(16) = 4102 ,

4624(10) = 4624 ,

1324(6) = 340 ,

1010(2) = 10 ,

1101(2) = 13 ,

4 +4 +4 +4

4 +4 +4 +4

4 +4 +4 +4

4 +4 +4 +4

0.593s, where they analyzed a total of 225352 cases.

39

52 + 56 + 53 + 51 + 57

26317(14) = 93905 ,

41277(14) = 156905 ,

67177(14) = 250005 ,

11346(11) = 16385 ,

23105(6) = 3281 ,

5 +5 +5 +5 +5

5 +5 +5 +5 +5

5 +5 +5 +5 +5

5 +5 +5 +5 +5

4.898s, where they analyzed a total of 3103928 cases

The solution with n = 6 digits and for b = 16, b = 15, . . . b = 2:

65 + 66 + 61 + 64 + 60 + 67

561407(9) = 335671 ,

84.037s, where they analyzed a total of 43912360 cases

The solutions with n = 7 digits and for b = 16, b = 15, . . . b = 2:

73 + 7 9 + 7 6 + 7 1 + 7 4 + 7 5 + 7 7

3961457(15) = 41314357 ,

1521327(9) = 840805 ,

1524117(9) = 842821 ,

1527424(9) = 845257 ,

1266643(8) = 355747 ,

7 +7 +7 +7 +7 +7 +7

7 +7 +7 +7 +7 +7 +7

7 +7 +7 +7 +7 +7 +7

7 +7 +7 +7 +7 +7 +7

24 : 53.587 mmss, where they analyzed a total of 633596120 cases

The solutions with n = 8 digits and for b = 16, b = 15, . . . b = 2:

83 + 8 9 + 8 1 + 8 6 + 8 0 + 8 0 + 8 1 + 8 6

39160016(12) = 134742546 ,

39160062(12) = 134742602 ,

76940369(12) = 271061505 ,

35161835(9) = 17105936 ,

8 +8 +8 +8 +8 +8 +8 +8

8 +8 +8 +8 +8 +8 +8 +8

8 +8 +8 +8 +8 +8 +8 +8

24 : 53.587 mmss, where they analyzed a total of 9280593352 cases

The solutions with n = 9 digits and for b = 14, b = 13, . . . b = 2 not exist,

where they analyzed a total of 37253987498 cases.

Programs

P2 (b) := k 0

40

n2

for j 0..b 1

D j nj

for d1 1..b 1

N1 d1 b

for d0 0..b 1

M Dd1 + Dd0

N N1 + d0

if N=M

sk (d1 d0 )

k k+1

return s

Programs for finding numbers N digits are similar. Program for finding inverse

Narcissistic numbers with 3 digits.

P3 (b) := k 0

n3

for j 0..b 1

D j nj

for j 2..n 1

B j bj

for d2 1..b 1

N2 d2 B2

for d1 0..b 1

N1 d1 b

for d0 0..b 1

M Dd2 + Dd1 + Dd0

N N2 + N1 + d 0

if N=M

sk (d2 d1 d0 )

k k+1

return s

Program call is P3 (9) = [(1 0 4)], where 31 + 30 + 34 = 104(9) = 85.

References

[1] M. Rumney, Digital Invariants, Recr. Math. Mag., 12, (1962), 68

[2] A. H. Beiler, Recreations in the Theory of Numbers, Dover, 1964

[3] J. S. Madachy, Narcissistic Numbers. Madachys Mathematical Recreations,

Dover, 1979, New York

41

Techniques, Results and Observations, J. Recr. Math., 14, (1981), 97108

[5] C. S. Ogilvy and J. T. Anderson, Excursions in number theory, Dover, 1988

[6] J. Roberts, The Lure of the Integers, Math. Assoc. Amer., 1992, Washington,

DC

[7] G. H. Hardy, A Mathematicians Apology, Cambridge University Press,

1993, New York

[8] C. A. Pickover, Keys to Infinity, Wiley, 1995, New York

[9] D. D. Spencer, Exploring number theory with microcomputers, Camelot

Pub. Co., 1995

[10] H. Heinz, Narcissistic Numbers,

September 1998

http://www.magic-squares.net/,

[11] C. C. Clawson, Mathematical Mysteries: The Beauty and Magic of Numbers, Perseus Books, 1996, Cambridge, Massachusetts

[12] C. A. Pickover, The Latest Gossip on Narcissistic Numbers, Chapter 88

in Wonders of Numbers: Adventures in Mathematics, Mind, and Meaning,

204205, Oxford University Press, 2001, Oxford, England

[13] R. K. Guy, Unsolved Problems in Number Theory, Springer-Verlag, 2004,

ed.3rd

[14] O. Cira and C. M. Cira, Narcissistic numbers, International Symposium

Research and Education in Innovation Era 3rd Edition, 2010, 197207,

Arad, Octomber, Aurel Vlaicu University of Arad

[15] M. Keith, Wild Narcissistic Numbers,

http://www.cadaeic.net/wild.htm

[16] C. Rivera, Problems & Puzzles: Puzzle 015-Narcissistic and Handsome

Primes, http://www.primepuzzles.net, May 2012

[17] N. J. A. Sloane, A005188/M0488, A003321/M5403, A010344, A010346,

A010348, A010350, A010353, A010354, A014576, A023052, A032799,

A046074, A101337, and A114904 in: The On-Line Encyclopedia of Integer

Sequences, http://oeis.org/, May 2012

[18] E. Fridman, Erichs Place: Whats Special About This Number?,

http://www2.stetson.edu/efriedma/index.html, May 2012

imbalanced datasets

Carmen T

erei

Marius Tomescu

Abstract

Any data set that exhibits an unequal distribution between its classes

can be considered imbalanced. Imbalance severely biases the learning process and badly affects the performances of the mining algorithms, because

often the classiffier has not enough samples of the minority class for proper

training. In the current paper we are making a comparative study on some

of the most popular classification algorithms for imbalanced datasets.

Mathematics Subject Classification: 68T05

Keywords: decision trees, support vector machines, undersampling, oversampling, cost-sensitive learning

Introduction

between its classes can be considered imbalanced. However, the common

understanding in the community is that imbalanced data correspond to

data sets exhibiting significant, and in some cases extreme, imbalances.

Specifically, this form of imbalance is referred to as a between-class imbalance; not uncommon are between-class imbalances on the order of 100:1,

1,000:1, and 10,000:1, where in each case, one class severely outrepresents

another[1]. Imbalances of this form are commonly referred to as intrinsic,

i.e., the imbalance is a direct result of the nature of the dataspace.

However, imbalanced data are not solely restricted to the intrinsic variety. Variable factors such as time and storage also give rise to data sets

that are imbalanced. Imbalances of this type are considered extrinsic, i.e.,

the imbalance is not directly related to the nature of the dataspace. Extrinsic imbalances are equally as interesting as their intrinsic counterparts

since it may very well occur that the dataspace from which an extrinsic

imbalanced data set is attained may not be imbalanced at all.

In addition to intrinsic and extrinsic imbalance, it is important to understand the difference between relative imbalance and imbalance due to

carmen terei@yahoo.com Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

tom uav@yahoo.com Department of Mathematics and Computer Science, Aurel Vlaicu

University of Arad, Romania

42

representative in domains where minority class examples are very limited,

i.e., where the target concept is rare. In this situation, the lack of representative data will make learning difficult regardless of the between-class

imbalance.

This, in fact, is the result of another form of imbalance, a within-class

imbalance, which concerns itself with the distribution of representative

data for subconcepts within a class.

The existence of within-class imbalances is closely intertwined with the

problem of small disjuncts, which has been shown to greatly depreciate

classification performance. The problem of small disjuncts can be understood as follows: a classifier will attempt to learn a concept by creating

multiple disjunct rules that describe the main concept. In the case of

homogeneous concepts, the classifier will generally create large disjuncts,

i.e., rules that cover a large portion (cluster) of examples pertaining to

the main concept.

In most cases, the imbalanced class problem is associated to binary

classification, but the multi-class problem often occurs and, since there

can be several minority classes, it is more difficult to solve [2][3].

Imbalance severely biases the learning process and badly affects the

performances of the mining algorithms, because often the classifier has not

enough samples of the minority class for proper training. The minority

class usually represents the most important concept to be learned, and it

is difficult to identify it since it might be associated with exceptional and

significant cases [4], or because the data acquisition of these examples is

costly [5].

Classification algorithms

A large number of approaches have been proposed to deal with the class

imbalance problem. These approaches can be categorized into two groups:

the internal approaches that create new algorithms or modify existing

ones to take the class-imbalance problem into consideration [6, 7, 8] and

external approaches that preprocess the data in order to diminish the

effect of their class imbalance.

Furthermore, cost-sensitive learning solutions incorporating both the

data (external) and algorithmic level (internal) approaches assume higher

misclassification costs for samples in the minority class and seek to minimize the high cost errors [9, 10, 11, 12].

There are more approaches, but they are not subject of the current

paper.

Internal approaches

1. Decision trees

Decision trees use simple knowledge representation to classify examples

into a finite number of classes. In a typical setting, the tree nodes represent

the attributes, the edges represent the possible values for a particular

43

attribute, and the leaves are assigned with class labels. Classifying a test

sample is straightforward once a decision tree has been constructed. An

object is classified by following paths from the root node through the tree

to a leaf, taking the edges corresponding to the values of attributes.

C4.5 is a decision tree generating algorithm. It induces classification

rules in the form of decision trees from a set of given examples. The

decision tree is constructed top-down using the normalized information

gain (difference in entropy) that results from choosing an attribute for

splitting the data. The attribute with the highest normalized information

gain is chosen to make the decision. The C4.5 algorithm then recurs on

the smaller sublists.

2. Support vector machines

Support vector machines (SVM) belong to a family of generalized linear models which achieves a classification model based on the linear combination of independent variables. The mapping function in SVM can

be either a classification function or a regression function. For classification, nonlinear kernel functions are often used to transform the input data

(inherently representing highly complex nonlinear relationships) to a high

dimensional feature space in which the input data becomes more separable

(i.e., linearly separable) compared to the original input space. Then, the

maximum-margin hyperplanes are constructed to optimally separate the

classes in the training data. The assumption is that the larger the margin or distance between these hyperplanes the better the generalization

performance of the classifier.

External approaches

The imbalanced data problem is quite usual in machine learning and data

mining applications as it appears in many real-world prediction tasks.

However, the techniques and concept of balancing the data prior to model

building is relatively new to many information systems researchers. A

wide variety of balancing techniques have been applied to data sets in

many areas such as medical diagnosis, classifiers for database marketing,

property refinance prediction, among others.

Preprocessing imbalanced datasets: resampling techniques

Sampling strategies are often used to overcome the class imbalance

problem.

It was empirically proved that applying a preprocessing step in order to balance the class distribution is usually an useful solution [13][14].

Furthermore, the main advantage of these techniques is that they are

independent of the underlying classifier.

Resampling techniques can be categorized into three groups or families:

1. Undersampling methods, which create a subset of the original

dataset by eliminating instances (usually majority class instances).

2. Oversampling methods, which create a superset of the original

dataset by replicating some instances or creating new instances from existing ones.

3. Hybrids methods, which combine both sampling approaches from

above.

44

originally developed by Chawla et al. (2002), generates synthetic minority

examples to be added to the original dataset. For each minority example,

its k nearest neighbors of the same class are found. Some of these nearest neighbors are then randomly selected according to the over-sampling

rate. A new synthetic example is generated along the line between the

minority example and every one of its selected nearest neighbors. This

process is repeated until the number of examples in all classes is roughly

equal to each other. Instead of replicating the existing instances, SMOTE

generates new synthetic minority class instances by interpolating between

several minority class examples that lie close together. It allows the classifiers to carve broader decision regions, which leads to more coverage of

the minority class.

Cost-sensitive learning

is another way to improve the performance of a classifier. Cost-sensitive

learning methods try to maximize a loss function associated with a data

set. These learning methods are motivated by finding that most real-world

applications do not have uniform costs for misclassifications.

Cost-sensitive learning takes into account the variable cost of a misclassification with respect to the different classes. In this case, a cost

matrix codifies the penalties C(i, j) of classifying examples of one class i

as a different one j.

Weighted LPSVM (wLPSVM) is a solution to the class imbalance

problem for LPSVM. It can be seen as a example of the weighting method.

The class imbalance problem for LPSVM is mainly caused by the imbalance force of the two-class training error. The idea of wLPSVM is: balancing the training error from two classes to obtain a separating plane

learning that is robust to the class imbalance. This training error balancing is achieved by assigning smaller weight to the training errors from the

majority class while assigning larger weight to that of the minority class.

The cost-sensitive C4.5 decision tree (C4.5CS) is a method to induce

cost-sensitive trees that seeks to minimize the number of high cost errors

and, as a consequence of that, leads to minimization of the total misclassification costs in most cases. The method changes the class distribution

such that the tree induced is in favor of the class with high weight/cost

and is less likely to commit errors with high cost. C4.5CS modifies the

weight of an instance proportional to the cost of misclassifying the class

to which the instance belonged, leaving the sum of all training instance

weights still equal to N . Let C(j) be the cost of misclassifying a class j

instance; the weight of a class j instance can be computed as

N

w(j) = C(j) P

C(i)N

i

i

P

such that the sum of all instance weights is j w(j)Nj = N .

45

A study on several datasets from various fields and different imbalance

ratios on which were applied preprocessing methods and cost-sensitive

learning algorithms has lead to some useful conclusions.

a. Study on the preprocessing methods

Some of the most representative techniques were compared, developing a ranking according to the performance obtained in each case.

The set of methods is composed by the following techniques: SMOTE,

SMOTE+ENN, Borderline-SMOTE (Border-SMOTE), ADASYN, SafeLevel-SMOTE (SL-SMOTE), SPIDER2 and DBSMOTE. The interpolations that are computed to generate new synthetic data were made considering the 5-nearest neighbors of minority class instances using the euclidean distance.

The conclusion was that SMOTE+ENN and SMOTE obtained the

highest performance, because of some possible reasons. The first one is

related to the addition of significant information within the minority class

examples by including new synthetic examples. The second reason is that

the more sophisticated the technique is, the less general it becomes for

the high number of benchmark problems selected for our study.

b. Study on the cost-sensitive learning algorithms

Three different approaches were used, namely the CS-Weighted, MetaCost, and the CostSensitive Classifier (CS-Classifier) from the Weka environment. In the first case, the base classifiers were modified usually by

weighting the instances of the dataset to take into account the a priori

probabilities, according to the number of samples in each class. In the

two latter cases, it was used an input cost-matrix defining C(+, ) = IR

and C(, +) = 1.

It has been appreciated that the CS-Weighted approach achieves the

highest rank overall. The MetaCost method obtains also a good average

for C4.5 and kNN, but it is outperformed by the CS-Classifier when SVM

is used.

The good behavior shown by introducing weights to the training examples can be explained by its simplicity, because the algorithm procedure

is maintained and is adapted to the imbalanced situation. Therefore,

it works similarly to an oversampling approach but without adding new

samples and complexity to the problem itself. On the other hand, the

MetaCost method follows a similar aim, therefore obtaining high quality

results.

Conclusions

datasets. The approaches which were studied were the datasets preprocessing methods and cost-sensitive learning algorithms. The aim was to

emphasize which ones lead to the best data classification results.

46

References

[1] H. He and X. Shen, A Ranked Subspace Learning Method for Gene

Expression Data Classification, Proc. Intl Conf. Artificial Intelligence, pp. 358-364, 2007.

[2] A. Fernandez, V. Lopez, M. Galar, M.J. del Jesus, F. Herrera,

Analysing the classification of imbalanced data-sets with multiple

classes: binarization techniques and ad-hoc approaches, KnowledgeBased Systems 42 (2013) 97110.

[3] M. Lin, K. Tang, X. Yao, Dynamic sampling approach to training

neural networks for multiclass imbalance classification, IEEE Transactions on Neural Networks and Learning Systems 24 (4) (2013)

647660.

[4] G.M. Weiss, Mining with rarity: a unifying framework, SIGKDD

Explorations 6 (1) (2004) 719.

[5] G.M. Weiss, Y. Tian, Maximizing classifier utility when there are

data acquisition and modeling costs, Data Mining and Knowledge

Discovery 17 (2) (2008) 253282.

[6] P. Ducange, B. Lazzerini, F. Marcelloni, Multi-objective genetic fuzzy

classifiers for imbalanced and cost-sensitive datasets, Soft Computing

14 (7) (2010) 713728.

[7] W. Lin, J.J. Chen, Class-imbalanced classifiers for high-dimensional

data, Briefings in Bioinformatics 14 (1) (2013) 1326.

[8] W. Zong, G.-B. Huang, Y. Chen, Weighted extreme learning machine

for imbalance learning, Neurocomputing 101 (2013) 229242.

[9] R. Batuwita, V. Palade, Class imbalance learning methods for support vector machines, in: H. He, Y. Ma (Eds.), Imbalanced Learning:

Foundations, Algorithms, and Applications, Wiley, 2013, pp. 8396.

[10] N. Garcia-Pedrajas, J. Perez-Rodriguez, M. Garcia-Pedrajas, D.

Ortiz-Boyer, C. Fyfe, Class imbalance methods for translation initiation site recognition in DNA sequences, Knowledge Based Systems

25 (1) (2012) 2234.

[11] Y. Sun, M.S. Kamel, A.K.C. Wong, Y. Wang, Cost-sensitive boosting

for classification of imbalanced data, Pattern Recognition 40 (12)

(2007) 3358 3378.

[12] Z.-H. Zhou, X.-Y. Liu, Training cost-sensitive neural networks with

methods addressing the class imbalance problem, IEEE Transactions

on Knowledge and Data Engineering 18 (1) (2006) 6377.

[13] R. Batuwita, V. Palade, Efficient resampling methods for training

support vector machines with imbalanced datasets, in: Proceedings of the 2010 International Joint Conference on Neural Networks

(IJCNN), 2010.

[14] A. Fernndez, M.J. del Jesus, F. Herrera, On the 2-tuples based genetic tuning performance for fuzzy rule based classification systems in

imbalanced data-sets, Information Sciences 180 (8) (2010) 12681291.

47

for imbalanced data

Carmen T

erei

Marius Tomescu

Abstract

The datasets that exhibit an unequal distribution between their classes

can be considered imbalanced. Imbalance severely biases the learning process and badly affects the performances of the mining algorithms. Conventional evaluation practice of using singular assessment criteria, such

as the overall accuracy or error rate, does not provide adequate information in the case of imbalanced learning. In the current paper we are

presenting some of the metrics which are dedicated for evaluating the data

classification for imbalanced datasets.

Mathematics Subject Classification: 68T05

Keywords: imbalanced data, confusion matrix, accuracy, Receiver Operating

Characteristic (ROC)

Introduction

As the research community continues to develop a greater number of intricate and promising imbalanced learning algorithms, it becomes a must to

have standardized evaluation metrics to properly assess the effectiveness

of such algorithms.

The measures of the quality of classification are built from a confusion

matrix, which records correctly and incorrectly recognized examples for

each class. The most used empirical measure, accuracy, does not distinguish between the number of correct labels of different classes, which in

the ambit of imbalanced problems may lead to erroneous conclusions. For

example a classifier that obtains an accuracy of 90% in a data-set with

a 90% of negative instances, might not be accurate if it does not cover

correctly any positive class instance.

Positive class

Negative class

Positive prediction

True Positive (TP)

False Positive (FP)

Negative prediction

False negative (FN)

True negative (TN)

Vlaicu University of Arad, Romania

tom uav@yahoo.com Department of Mathematics and Computer Science, Aurel Vlaicu

University of Arad, Romania

48

in this situation are considered. Two common measures, sensitivity and

specificity, approximate the probability of the positive (negative) label

being true. In other words, they assess the effectiveness of the algorithm

on a single class.

Numerical metrics

Accuracy, Recall, Precision, F-measure, kappa, ACU and some other new

proposed measures like Informedness and Markedness are examples of

different evaluation measures.

Depending on the problem and the field of application one measure

could be more suitable than another. While in the behavioral sciences,

Specificity and Sensitivity are commonly used, in the medical sciences,

ROC analysis is a standard for evaluation. On the other hand, in the Information Retrieval community and fraud detection, Recall, Precision and

F-measure are considered appropriate measures for testing effectiveness.

In a learning design strategy, the best rule for the specific application

will be the one that get the optimal performance for the chosen measure.

This problem is particularly important in those applications where the

instances of a class (the majority one) heavily out- number the instances of

the other (the minority) class and it is costly to misclassify samples from

the minority class. For example in information retrieval, nontechnical

losses in power utilities or medical diagnosis.

For example, in a decision tree the pruning criterion is usually the

classification error, which can remove branches related with the minority

class. In back- propagation neural networks, the expected gradient vector

length is proportional to the class size, and so the gradient vector is dominated by the prevalent class and consequently the weights are determined

by this class.

SVMs are thought to be more robust to the class imbalance problem

since they use only a few support vectors to calculate region boundaries.

However, in a two class problem, the boundaries are determined by the

prevalent class, since the algorithm tries to find the largest margin and the

minimum error. A different approach is taken in one-class learning, for

example one class SVM, where the model is created based on the samples

of only one of the classes. In [21] the optimality of one-class SVMs over

two-class SVM classifiers is demonstrated for some important imbalanced

problems.

In most of the approaches that deal with an imbalanced problem, the

idea is to adapt the classifiers that have good Accuracy in balanced domains. A variety of ways of doing this have been proposed: changing class

distributions, incorporating costs3 in decision making, and using alternative performance metrics instead of Accuracy in the learning process with

the standard algorithms.

The evaluation criteria is a key factor in assessing the classification

performance and guiding the classifier modeling. In a two-class problem,

49

the confusion matrix records the results of correctly and incorrectly recognized examples of each class.

Traditionally, the accuracy rate has been the most commonly used empirical measure. However, in the framework of imbalanced datasets, accuracy is no longer a proper measure, since it does not distinguish between

the number of correctly classified examples of different classes. Hence, it

may lead to erroneous conclusions, i.e., a classifier achieving an accuracy

of 90% in a dataset with an IR value of 9 is not accurate if it classifies all

examples as negatives.

TP + TN

Acc =

TP + FN + FP + TN

In imbalanced domains, the evaluation of the classifiers performance

must be carried out using specific metrics in order to take into account

the class distribution. Concretely, we can obtain four metrics to measure the classification performance of both, positive and negative, classes

independently:

True positive rate: T Prate =

instances correctly classified.

TP

T P +F N

instances correctly classified.

TN

F P +T N

instances misclassified.

FP

F P +T N

instances misclassified.

FN

T P +F N

mean of the

rtrue rates, which can be defined as:

TP

TN

GM =

TP + FN FP + TN

This metric attempts to maximize the accuracy on each of the two

classes with a good balance, being a performance metric that correlates

both objectives. However, due to this symmetric nature of the distribution

of the geometric mean over TPrate (sensitivity) and the TNrate (specificity), it is hard to contrast different models according to their precision

on each class.

Another significant performance metric that is commonly used is the

F-measure:

(1 + 2 )(P P V T Prate )

Fm =

2 P P V + T Prate

TP

PPV =

TP + TP + FP

According to the previous comments, some authors try to propose

several measures for imbalanced domains in order to be able to obtain as

much information as possible about the contribution of each class to the

final performance and to take into account the IR of the dataset as an

indication of its difficulty.

For example, in [10,14] the Adjusted G-mean is proposed. This measure is designed towards obtaining the highest sensitivity (TPrate) without decreasing too much the specificity (TNrate). This fact is measured

50

with respect to the original model, i.e. the original classifier without addressing the class imbalance problem.

GM + T Nrate (F P + T N )

AGM =

; If T Prate > 0

1 + FP + TN

AGM = 0; If T Prate = 0

Additionally, in [54] the authors presented a simple performance metric

called Dominance, which is aimed to point out the dominance or prevalence relationship between the positive class and the negative class, in the

range [1, +1]. Furthermore, it can be used as a visual tool to analyze

the behavior of a classifier on a 2-D space from the join perspective of

global precision (Y-axis) and dominance (X-axis).

Dom = T Prate T Nrate

The same authors, using the previous concept of dominance, Index

of Balanced Accuracy (IBA). IBA weights a performance measure, that

aims to make it more sensitive for imbalanced domains. The weighting

factor favors those results with moderately better classification rates on

the minority class. IBA is formulated as follows:

IBA (M ) = (1 + Dom)M

where (1 + Dom) is the weighting factor and M represents a performance metric. The objective is to moderately favor the classification

models with higher prediction rate on the minority class (without underestimating the relevance of the majority class) by means of a weighted

function of any plain performance evaluation measure.

Graphical metrics

A well-known approach to unify these measures and to produce an evaluation criteria is to use the Receiver Operating Characteristic (ROC)

graphic. This graphic allows the visualization of the trade-off between

the benefits (TPrate) and costs (FPrate), as it evidences that any classifier cannot increase the number of true positives without also increasing

the false positives.

ROC curves, graph true positive rates on the y-axis vs. the false positive rates on the x-axis. The resulting curve illustrates the trade-off between detection rate and false alarm rate. The ROC curve illustrates the

performance of a classifier across the complete range of possible decision

thresholds, and accordingly does not assume any particular misclassification costs or class prior probabilities.

For a single numeric measure, the area under the ROC curve (AUC)

is widely used, providing a general idea of the predictive potential of the

classifier. A higher AUC is better, as it indicates that the classifier, across

the entire possible range of decision thresholds, has a higher true positive

rate. The AUC is the performance metric used for this study. Provost

and Fawcett give an extensive overview of ROC curves and their potential

use for creating optimal classifiers.

The Area Under the ROC Curve (AUC)corresponds to the probability

of correctly identifying which one of the two stimuli is noise and which one

is signal plus noise. The AUC provides a single measure of a classifiers

51

measure is computed just by obtaining the area of the graphic:

1 + T Prate + F Prate

AU C =

2

According to the authors, the main advantage of this type of methods

resides in their ability to depict the trade-offs between evaluation aspects

in a multidimensional space rather than reducing these aspects to an arbitrarily chosen (and often biased) single scalar measure. In particular, they

present a review of several representation mechanisms emphasizing the

best scenario for their use; for example, in imbalanced domains, when we

are interested in the positive class, it is recommended the use of precisionrecall graphs [36]. Furthermore, the expected cost or profit of each model

might be analyzed using cost curves, lift and ROI graphs.

Conclusions

In this paper we presented some of the techiques which are adapted for

evaluating the quality of the algorithms for the classification of imbalanced

datasets. There are several numerical metrics and also some graphical

metrics.

References

[1] H. He and X. Shen, A Ranked Subspace Learning Method for Gene

Expression Data Classification, Proc. Intl Conf. Artificial Intelligence, pp. 358-364, 2007.

[2] A. Fernandez, V. Lopez, M. Galar, M.J. del Jesus, F. Herrera,

Analysing the classification of imbalanced data-sets with multiple

classes: binarization techniques and ad-hoc approaches, KnowledgeBased Systems 42 (2013) 97110.

[3] M. Lin, K. Tang, X. Yao, Dynamic sampling approach to training

neural networks for multiclass imbalance classification, IEEE Transactions on Neural Networks and Learning Systems 24 (4) (2013)

647660.

[4] G.M. Weiss, Mining with rarity: a unifying framework, SIGKDD

Explorations 6 (1) (2004) 719.

[5] G.M. Weiss, Y. Tian, Maximizing classifier utility when there are

data acquisition and modeling costs, Data Mining and Knowledge

Discovery 17 (2) (2008) 253282.

[6] P. Ducange, B. Lazzerini, F. Marcelloni, Multi-objective genetic fuzzy

classifiers for imbalanced and cost-sensitive datasets, Soft Computing

14 (7) (2010) 713728.

[7] W. Lin, J.J. Chen, Class-imbalanced classifiers for high-dimensional

data, Briefings in Bioinformatics 14 (1) (2013) 1326.

[8] W. Zong, G.-B. Huang, Y. Chen, Weighted extreme learning machine

for imbalance learning, Neurocomputing 101 (2013) 229242.

52

[9] R. Batuwita, V. Palade, Class imbalance learning methods for support vector machines, in: H. He, Y. Ma (Eds.), Imbalanced Learning:

Foundations, Algorithms, and Applications, Wiley, 2013, pp. 8396.

[10] N. Garcia-Pedrajas, J. Perez-Rodriguez, M. Garcia-Pedrajas, D.

Ortiz-Boyer, C. Fyfe, Class imbalance methods for translation initiation site recognition in DNA sequences, Knowledge Based Systems

25 (1) (2012) 2234.

[11] Y. Sun, M.S. Kamel, A.K.C. Wong, Y. Wang, Cost-sensitive boosting

for classification of imbalanced data, Pattern Recognition 40 (12)

(2007) 3358 3378.

[12] Z.-H. Zhou, X.-Y. Liu, Training cost-sensitive neural networks with

methods addressing the class imbalance problem, IEEE Transactions

on Knowledge and Data Engineering 18 (1) (2006) 6377.

[13] N.V. Chawla, A. Lazarevic, L.O. Hall, K.W. Bowyer, SMOTEBoost:

Improving prediction of the minority class in boosting, in: Proceedings of 7th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD03), 2003, pp. 107119.

[14] R. Batuwita, V. Palade, Efficient resampling methods for training

support vector machines with imbalanced datasets, in: Proceedings of the 2010 International Joint Conference on Neural Networks

(IJCNN), 2010.

[15] A. Fernandez, M.J. del Jesus, F. Herrera, On the 2-tuples based genetic tuning performance for fuzzy rule based classification systems in

imbalanced data-sets, Information Sciences 180 (8) (2010) 12681291.

[16] G.E.A.P.A.Batista, R.C.Prati, M.C.Monard, A study of the behaviour of several methods for balancing machine learning training

data, SIGKDD Explorations 6 (1)(2004)2029.

53

Mariana Nagy

Negrusa Adina

Burca Valentin

Abstract

Financial statements are essential on capital investment decision, especially in the new context of a globalized world, increasing international

capital market listing, or amplified commercial relation among national

economies, and the promising international process of accounting convergence. Our study aim to analyze the closeness of different IFRS adoption

strategies, by creating four clusters, based on the level of IFRS adoption

referring to consolidated financial statements, individual financial statements and simplified financial statements accounting regulation. The results reveal small differences between the clusters determined, leading to

a promising future of the recent IASB projects of continuous improvement

of existing standards and for new standards.

Jel classification:

Keywords:

alization.

M21, M41.

Introduction

modification in financial position within financial statements has raised vivid

debate among the researchers, especially in the past decades characterized by a

strong will of reducing international accounting differences that affected seriously

decision-making process. Most recent ambitious project of international accounting convergence, conducted by IASB in cooperation with FASB, supported by

international professional organizations and local government institutions, has

underlined not just the opportunity of improving financial information quality,

but also has shown chances of cost reduction at firm-level and country-level as

well.

Though, the process of international accounting convergence seems to take a

longer time than expected, the signals are positive as nowadays there are more

than 100 jurisdictions that decided to adopt IFRS opting for different strategies

well known at international level. Once the capital markets have opened on international listing and foreign investments have increased exponentially, especially

first.author@uav.ro PhD prof. Aurel Vlaicu University of Arad Department of Mathematics and Computer Science, Aurel Vlaicu University of Arad, Romania

second.author@uvt.ro PhD, assoc. prof., Babe

s Bolyai University, Cluj-Napoca

third.author@uvt.ro PhD candidate, West University of Timisoara Department of Mathematics, West University of Timisoara, Romania

54

in the case of the emerging and economies in transition, the incentives offered for

IFRS adoption has determined investors and international financial institutions

to make pressure for IFRS adoption, against the political factor option (Burca

& Cilan, 2013).

Unfortunately, even within countries with similar accounting systems, persist

accounting differences caused by various factors describing the environment in

all its aspects. The problem of overt and covert options permitted by IFRSs

and the ambiguity on defining and utilizing various conceptual terms has led a

more relevant financial information than the previous period of local GAAPs,

but hasnt solved completely the problem of opportune accounting strategies

(Nobes, 2008; Doupnik & Perrera, 2012).

On this article we attempt to make a classification of jurisdictions worldwide

considering different strategies of IFRS adoption used, as the adoption described

a gradual transition from local GAAP to IFRS. The global positive perception

for IFRS adoption is valid, being reiterated by all accounting regulators. But

there still persist reluctance on implementing IFRS for all financial statements

for most of the jurisdictions. We subscribe to Christensen (2012) opinion who

sustain that a cause may be the inconsistencies around benefits and costs of

IFRS adoption revealed by numerous recent studies.

Literature review

The step towards a unique set of international core standards has been promoted

fervently by capital market actors, as the comparability and value relevance of

the financial information is essential investment decision. The option for IFRS

adoption is recommended in the light of the results of recent studies that envisage

a more accurate reporting framework drawn by IFRSs, that improve significantly

the quality of financial information (Daske et. al., 2008; Ramana & Sletten,

2009; Chen et. al., 2010; Barth et. al., 2012; Daske et. al., 2013), leading to

macroeconomic and microeconomic positive effects (Daske et. al., 2013; Banker

et. al., 2014; Biddle et. al., 2013).

Even if there were outlined serious compatibility issues between local accounting system and IFRS provisions, recent studies revealed that fundamental

in the transition process to IFRS regulation is the power of will of change. Moreover, the discussion around the overt and covert accounting policy options seems

to be more complex, underlining the trade-off between the preparers and the user

of the financial information. Essential on this debate is the power of intention

when designing the accounting policies (Fields et. al., 2001; Jianu, 2012).

It is less important what financial statement depict a faithful representation

of the economic reality, as this can be easily set by the financial system, in the

limits of efficiency (Ball & Shivakumar, 2005; Nobes & Parker, 2008). Important

is that the mechanism and tools of financial reporting to disclose any doubt of

accounting manipulation. Here the role of the institutional framework of a jurisdiction, and complementary a consolidated mechanism of corporate governance,

have to bring their contribution to the proper enforcement of the new accounting

regulation (Samaresekera et. al., 2012; Christensen et. al., 2013).

Most of negative perception translated in reluctance to IFRS adoption of

some firms was mainly determined by the lack of prior studies, or insignificant

reporting incentives provided by an emerging capital market. On this context it

is essential that accounting standard-setters and local enforcement institutions

to pay attention to the role of reporting incentives (especially the market-driven

ones), cause the quality of the accounting standards does not necessary traduce

into qualitative financial reporting (Burgstahler et. al.2006; Jayaraman & Verdi,

2014).

The option for IFRS standards, but not for US GAAP, is not only a matter

of a economic cost-benefits analysis, but also, a matter of politics of accounting

55

accounting regulation with IFRS provisions, explaining these strategies by two

factors: countrys potential influence on IASB decisions and the proximity of

the existing accounting system to the ones of the political powers at the IASB.

Nobes (2011) identify six ways of IFRS implementation:

full adoption of IFRS, when the jurisdiction use the most recent versions

of IFRSs, without passing the content of the standards though an endorsement process;

insertion of IFRSs into law, which differs from first way by delaying the

date of use of the new or revised standards;

endorsing IFRS, which involves detailed scrutiny of all IFRS standards,

leading many times to several carve-outs, altering the content quality of

the original IFRS standard;

full convergence with IFRS, with expressed intention of full compliance,

which imply IFRS insertion into national regulation with several amendments leading to textual changes, several carve-outs, and early adoption

banning;

adapting IFRS, is the case of classic accounting harmonization, when a

jurisdiction use IFRSs as a source of regulation;

allowing IFRS, is the way jurisdictions preserve their power of accounting

regulation for a restricted category of companies, as most of jurisdiction

do currently when discussing potential IFRS for SMEs adoption.

IFRS are currently used in the consolidated financial statements in IFRS

of over 100 jurisdictions .Adopting IFRS takes many forms. While some states

choose to implement the full incorporation of IFRS into national legislation

(Trinidad and Tobago -1988 , Bosnia - Herzegovina 1995 Vietnam - 2002 Bahamas 2007); or have successfully completed harmonization projects of local accounting standards with the international standards ( China - 2006-2010 Algeria

, India , Indonesia - 2012) most jurisdictions choose adopting IFRS in its original

form , by limiting the applying sphere only to consolidated financial statements

of listed companies . But currently, there are also jurisdictions (Cameroon ,

Congo , Senegal - 2014 , Bolivia , Colombia -2015, US- 2016 , Saudi Arabia 2017) that did not appeal to the adoption of IFRS (around 15% of jurisdictions

worldwide) but many of them have expressed their desire to conduct a future

project for implementing IFRS .

Methodological research

trend on the next future regarding IFRS adoption worldwide. The sample used

consist of 57 jurisdiction, which can be split by region as the chart below show.

The information regarding regulation of financial reporting according to

IFRS provisions are collected from multiple sources reminding here PWC (2013)

study ,,IFRS adoption by country, Deloitte website dedicated for IFRS concerns and IASB website relating jurisdictions profiles.

Making abstraction of the endorsement process used in case of some jurisdiction, as is the case of EU countries, the jurisdictions will be classified based on

a score variable designed to describe the strategies of IFRS adoption mentioned

above. The sample countries are first analyzed based on the treatment for IFRS

use for the following groups of companies: are first classified in three groups:

listed foreign companies, listed domestic companies and unlisted companies and

SMEs. The scor variable is calculated considering different multiples to be used

for the type of companies, the level of IFRS use on consolidated and statutory

56

financial statement as well is order to underline the reluctance of most developed jurisdictions to apply IFRS for all financial statements. For the treatment

of prohibiting IFRS we consider the value 1, as for the treatment of permitting

IFRS we use 2 and for the treatment of mandatory use of IFRS (value m ) there

is allocated value 3.

The classification is realized using the k-means cluster analysis technique

which have as variables the scores for each type of financial statement. Leuz

(2010) has achieved to classify 49 countries into a small number of clusters considering a suite of factors concerning institutional characteristics, such as legal

origin, disclosure requirements rule of law etc. Nanda & Wysocki (2011) study

the causal relation between societal trust and firms voluntary and regulated

financial reporting and disclosure quality. Later Nanda & Wysocki (2013) used

the same method of clustering, this time to analyze the relation between societal

trust and firms financial transparency, and how firms external capital demand

affects this relation.

In our analysis we will consider the number of necessary clusters to be identified by analyzing graphically the positioning of all jurisdictions on a 3D graph

describing all three above mentioned scores to be used (Field, 2005). The basic

rationale of our approach is that all jurisdictions included on our sample different

more or less on the level IFRS is allowed or even mandatory to use on preparing

financial statements. The evidence reveal a gradual IFRS adoption showing that

on first stage the consolidated financial statements only are prepared according

to IFRS.

Representation by type of financial statements regulation

On the second stage, the national regulators decide the extent of IFRS use

to preparation of individual financial statements. This step depends strongly on

how strong is the relation between accounting policies and fiscal rules, as the

continental accounting systems are well known for fiscal economies obtained with

the condition to comply with a prescribed formula of accounting registration of

different types of economic transactions (Lamb et. al.,1998).

The last stage is supposed to make the transition to IFRS of the small and

medium enterprises, too. On this direction IASB has designed a simplified set

of accounting standards, meant to respond properly to the lower demand of

financial information implied. This way they have published the IFRS for SMEs

57

that have already decided IFRS adoption for listed companies. But, till now

there is long debate around the opportunity of this decision. For instance, the EU

community has given great attention to this subject, even revising the European

directives regulating accounting treatments.

Representation by type of company group

by type of financial statements regulation and by type of company analyzed as

well. The general conclusion is that jurisdictions expressed their will to adopt

IFRS, but just for consolidated financial accounts as they are not used on fiscal

purpose, explained also by the negative strong correlation of about -0.288.

Indeed, there is seen on last years a more clear direction towards extent of

IFRS use to individual financial statements, too. The statistics reveal a relatively

high level of IFRS adoption, but with higher variation within the sample, but

the focus for mandatory IFRS adoption still continue to be in case of listed

companies. Unlisted companies are rather permitted to use IFRS on statutory

annual accounts, but just as a second set of financial statements which lead to

higher costs of preparation implied by the reconciliation costs registered. This

way, the voluntary adoption is really low, even if the literature confirmed higher

benefits of IFRS voluntary adoption than in case of mandatory IFRS adoption.

But this trend can be mainly justified by the pressure exerted by investors

and international financial institutions in case of developed countries, as they

already have a solid knowledge in financial accounting regulation, with a strong

financial education provided by well know universities, and a powerful accounting research tradition. But, jurisdictions perception on IFRS adoption depends

especially on the ratio between similarities and differences between the local

GAAP and IFRS standards.

Contrary, the underdeveloped economies and economies in transition, the

solution of fully IFRS adoption is more reliable as this would lead to significant

government cost reduction through the deregulation process. Here potential

issues can raise around the quality of training the professional on the IFRS hot

topics, such countries being forced more or less to access consulting services

provided by accounting and auditing international offices, such as the Big4

The same graph illustrate a persistent visible reluctance of jurisdictions towards IFRSs for SMEs adoption. But this reluctance is not valid only for the

political factor, being shown by managers of SMEs as well. In the case of SMEs

there is more visible the cultural component of managers behavior regarding

financial reporting strategies. If the fiscal factor remain a constant determinant

58

59

is evidence of a stronger correlation between score of IFRS for SMEs adoption1

and level of uncertainty avoidance perceived.

Proceeding to clustering the sample, we have obtained four slightly equal

clusters. They are different, especially based on the variations of treatment

regarding IFRS adoption for consolidated accounts preparation and treatment

regarding IFRS adoption for simplified accounts preparation:

the first cluster is dominated by the highest level of IFRS adoption, in

case of all types of companies and all levels of financial reporting. This

can be explained by the emergent capital markets, except the case of New

Zeeland which follow the way of IFRS cause of its common law origins;

the second cluster group jurisdictions which prefer IFRS adoption especially for the foreign listed companies, in order to support potential foreign

investor, leading to a growing market capitalization;

the third cluster is mainly comprised by EU community jurisdictions,

which are affected by the IAS regulation (1606/2002) which obliged listed

companies to report under the IFRS provisions the consolidated accounts.;

the forth cluster contains countries that rather permit IFRS use than decide for mandatory adoption in most cases, especially in the case of unlisted

companies and SMEs

There are not major differences between the four clusters, which means that

all jurisdictions tend to consider IFRS adoption as a viable scenario for future

1 Even if we refer to IFRS for SMEs adoption as the scenario in case of SMEs, there are

cases which still use complete version of IFRSs for SMEs, probably till the debate around the

opportunity for IFRS for SMEs adoption will end in favor;

accounting regulation. The differences consist of the timeframe of the transition process, the date of implementation, the differences of level of local GAAP

harmonization with IFRS etc. All the clusters describe one general trend of

IFRS adoption which predict a long way to fully IFRS adoption. We tend to

even think that this objective is impossible. But the interest for IFRS philosophy

and rationale still remain high among the regulators. Thus, even if the process of

international accounting convergence will not find a reliable solution for international accounting systems uniformity, the process of international harmonization

will continue.

Conclusions

It is obvious that IFRS represent the global solution for a new era of an international accounting language. Just that the uncertainty for new of most

jurisdictions is visible and the more flexible accounting philosophy promoted by

IASB means the professionals rationale to become the center of the entire financial reporting process. Unfortunately, not all the time the preparers of the

annual accounts prove to be honest and sincere, often making use of various creative accounting techniques, forced by economic context and shared life values.

Adding to this reality the fact that, on an international capital market and a

globalized economy, the international financial analyses are of high importance

on decision-making, the financial information is disclosed based on different accounting regulations, the quality of financial information decrease drastically.

If the contracting theory cant be solved by specific regulation, just through

an open negotiation language establishing the incentives area, the accounting

differences were significantly reduced by the international accounting harmonization process, and after by the more ambitious international accounting convergence efforts.

The evidence confirm as a general accepted solution worldwide the IFRS

adoption on a gradual process of transition. Unfortunately, the reality reveal

just a superficial success of the accounting convergence success as there is long

way till all jurisdictions will decide to extend IFRS use not just in purpose of

preparing the consolidated accounts. This is because of the reluctance of some

60

players who are still not sure about the impact of IFRS quality on the economy

and on the profound reform of the society.

We have revealed that the jurisdiction, even if differ based on the strategies

of IFRS adoption, eventually have built a positive perception about IFRS quality

and the necessity of international accounting practices comparability.

The complexity of the accounting convergence process explain the long timeframe necessary for full compliance with IFRS at a generalized level. The odds

will be in favor of IASB project till the political factor will support IASB work.

Otherwise, the limited legitimacy of IASB will not be enough for a proper IFRS

implementation. We underline the fact that IFRS accounting quality does not

transpose mandatory into qualitative accounting practices. There is necessary

a continuous relation of communication between managers, shareholders and

stakeholders as well, in order to build the bases of a confident business model

which can be financed at a lower cost of capital, registering high rate of returns.

It is obvious that nobody expect that IFRS provisions will solve entire problems raise along the financial reporting supply chain. That is why preparers and

users as well have to understand that their contribution to financial disclosures

optimization is essential, especially through the voluntary financial reporting

component.

References

[1] Banker R.D., Huang R., Li Y. (2014), Do Accounting Standards Matter for

Firm Productivity ? Evidence from Mandatory IFRS Adoption, available

on http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2429121;

[2] Barth M.E., Landsman W.R., Lang M.H., Williams C.D. (2012), Are IFRSbased and US GAAP-based Accounting Amounts Comparable, Journal of

Accounting and Economics, vol. 54, issue 1;.

[3] Biddle G.C., Callahan C.N., Hong H.A., Knowles R.L. (2013), Do Adoptions of International Financial Reporting Standards Enhance Capital Investment Efficiency?, available on http://papers.ssrn.com/sol3/papers.

cfm?abstract_id=2353693;

[4] Burca V., Cilan T.F. (2013), Multinationals Influence in Decision to Adopt

IAS/IFRS, Annals of the University of Petrosani- Economics, vol. 13, issue

1;

[5] Burgstahler D.C., Hail L., Leuz C. (2006), The Importance of Reporting

Incentives: Earnings Management in European Private and Public Firms,

The accounting Review, vol. 81, issue 5;

[6] Chen H., Tang Q., Jiang Y., Lin Z. (2010), The Role of International

Financial Reporting Standards in Accounting Quality: Evidence from the

European Union, Journal of International Financial Management & Accounting, vol. 21, issue 3;

[7] Christensen H. (2012), Why Do Firms Rarely Adopt IFRS Voluntarily?

Academics Find Significant Benefits and the Costs Appear to be Low, Review of Accounting Studies, vol. 17, issue 3;

[8] Christensen H.B., Hail L., Leuz C (2013), Mandatory IFRS Reporting and

Changes in Enforcement, Journal of Accounting and Economics, vol. 5,

issue 2-3;

[9] Daske H., Hail L., Laux C., Verdi R. (2008), Mandatory IFRS Reporting

Around the World: Early Evidence on the Economic Consequences, Journal

of Accounting Research, vol. 46, issue 5;

[10] Daske H., Hail L., Laux C., Verdi R. (2013), Adopting a Label: Heterogeneity in the Economic Consequences Around IAS/IFRS Adoptions, Journal

of Accounting Research, vol. 51, issue 3;

61

[11] Doupnik T., Perera H. (2012), International Accounting, McGraw Hill Education;

[12] Field A. (2005), Discovering statistics using SPSS, 3th edition, Sage Printing House, London;

[13] Fields T.D., Lys T.Z., Vincent L. (2001), Empirical research on accounting

choice, Journal of Accounting and Economics, vol. 31, issues 1-3;

[14] Jayaraman S., Verdi R.S. (2014), Are Reporting Incentives and Accounting

Standards Substitutes or Complements in Achieving Accounting Comparability, available on http://papers.ssrn.com/sol3/papers.cfm?abstract_

id=2428263;

[15] Lamb, M., Nobes, C., Roberts, A. (1998), International Variations in the

Connections Between Tax and Financial Reporting, Accounting and Business Research, vol. 28, issue 3;

[16] Leuz C. (2010), Different Approaches to Corporate Reporting Regulation:

How Jurisdictions Differ and Why, Accounting and Business Research, vol.

40, issue 3;

[17] Nanda D., Wysocki P. (2011), The Relation between Trust and Accounting Quality, available on http://management.bu.edu/files/2011/02/

Nanda-Wysocki-2011.pdf;

[18] Nobes C., Parker R. (2008), Comparative Financial Accounting, Prentice

Hall, London;

[19] Nobes C.,

2011,

International variations in IFRS adoption

and practice,

research report,

ACCA, London,

available on

http://www.accaglobal.com/content/dam/acca/global/PDF-technical/

financial-reporting/rr-124-001.pdf;

[20] PWC (2013), IFRS adoption by country, available on www.pwc.com;

[21] Ramana K. (2011), The International Politics of IFRS Harmonization,

available on http://hbswk.hbs.edu/item/6776.html ;

[22] Samaresekera N., Chang M., Tarca A. (2011), IFRS and accounting quality:

The impact of enforcement, http://papers.ssrn.com/sol3/papers.cfm?

abstract_id=2183061.

62

fields

Pastorel Gaspar

Lorena Popa

Abstract

In [3] we started a study of multivariate second order random distribution fields (m.s.o.r.d.f). This note is devoted to ordinary, modular and tensor series representations and some relation between these for m.s.o.r.d.f.

Finally a connection of series representation of m.s.o.r.d.f. with the correlation distribution is given.

Mathematics Subject Classification:

Keywords: random distribution field, series representation

Introduction

stochastic processes and random fields. Integral representations appear

often in the case of stationary random fields (see [6] and the literature

cited there). In [3] the authors extended the study of infinite dimensional

random fields to the setting of random distribution fields and stochastic

mappings, while [4] is devoted to the stationarity and stationarily cross

correlatedness including integral representations.

Series representations for stochastic processes appeared first in [7] and [8]

and then in [9] and [1]. This study was continued in [5] and [6]. In this

paper we consider the three types of series representations presented in

[6, Sec. 4.6] in the extended framework of [3] and [4]. Hence, we may use

to a significant extent concepts and notations from [6], [3] and [4].

So (, A , ) will be a probability space and H a separable Hilbert space,

for which B(H) and C1 (H) mean the Banach algebra of all continuous

linear operators on H, respectively the ideal of trace class operators from

B(H). The space L20 (, H) of second order H-valued random variables on

of zero mean will be denoted by H . Its elements will be also called

multivariate second order random variables, or briefly m.s.o.r.v. It will be

endowed (see [6, Sec. 1.3]) with a normal Hilbert B(H)-module structure,

pastorel.gaspar@uav.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

lorena.popa@yahoo.com Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

63

having a C1 (H)-valued gramian

Z

h() k()d(),

[h, k]H =

64

h, k H ,

tr[h, k]H , h, k H , is the natural scalar product making H a Hilbert

space. As such it can be identified (see also [6, Sec. 2.1]) with the Hilbert

space tensor product H L20 (), where L20 () : = L20 (, C). The norm on

H will be consequently expressed by

1/2

1/2

hH.

H -valued functions on Rd , but more generally as H -valued distributions

on Rd , then we consider D(Rd ) = Dd , the space of test functions in the

theory of distributions instead of the time parameter set of the field. Such

an object will be called m.s.o. random distribution field (m.s.o.r.d.f.).

multivariate second order random fields as were given in [6, Sec. 4.6]

to the setting treated in [3] and [4] of multivariate second order random

distribution fields.

A family of elements {h }I of H of norm 1 is said to be gramian

orthonormal if [h , h ]2 = [h , h ] for each I and [h , h ] = 0, for

6= I. A maximal gramian orthonormal family will be called gramian

orthonormal basis.

The following lemma (see [5]) is also useful in what follows.

Lemma 2.1 For a subset H H of the normal Hilbert module H , we

have that the submodule generated by it, S(H), is separable if the Hilbert

space generated by it, Sp(H), is separable.

We shall refer here to three such ways to express a multivariate second

order random distribution field U = {U }Dd . Namely we say that

U admits an ordinary series representation, briefly o.s.r., if there

exists a sequence of scalar distributions {uj }jN and a sequence

{hj }jN of elements from L20 [, H] = : H such that

U () =

uj ()hj () ;

(2.1)

j=1

U admits a modular series representation, briefly m.s.r., if there exists a sequence of B(H) - valued distributions {uj }jN and a sequence

of functions {hj }jN from L20 [, H] = H such that

U () =

uj ()hj () ;

j=1

(2.2)

65

Dd , U is the sum of a series of elementary tensors from Dd0 (H )

written in the form Dd0 (H) L20 (), i.e. there exists a sequence of

H-valued distributions {uj }jN and a sequence of functions {fj }jN

from L20 () such that for each Dd

U () =

uj ()fj () ;

(2.3)

j=1

We notice that it is immediate that a series representation of U of type

(2.1), implies a modular series representation, i.e. of type (2.2), if we put

uj () = uj ()IH , Dd .

Proposition 2.2 Suppose that the m.s.o.r.d.f. U has one of the above

three representations, where the systems {hj }jN are orthonormal in H

as a Hilbert space in the case of o.s.r., gramian orthonormal in H as

a normal Hilbert B(H)-module in the case of m.s.r. and {fj }jN are

orthonormal in L20 () in the case of t.s.r. Then the corresponding distributions forming the systems {uj }jN , {uj }jN , {uj }jN are respectively

uniquely determined.

Proof. We shall restrain ourselves to the last case. Suppose U has a

series representation (2.3) with an orthonormal basis {fj }jN L20 () and

P

U () =

vj ()fj () ; Dd , is another tensor representation

j=1

#

"

X

0 = fl h,

uj () vj () fj = h ul () vl () , Dd , l N.

j=1

Some

characterizations of the series representability of m.s.o.r.d.f. are contained

in

Theorem 2.3 Let {U }Dd be a m.s.o.r.d.f. The following characterizations hold:

(i) U has a representation of type (2.1) iff its vector time domain H(U )

is separable;

(ii) U has a representation of the form (2.2) iff its modular time domain

HU is separable;

(iii) U has a representation of the form (2.2) iff U has a tensor representation as a series, i.e. of the form (2.3).

Proof. It suffices to show (ii) since (i) can be deduced from (ii) for

H = C. Suppose that representation (2.2) holds for U . It results then

easy that HU is contained in the Hilbert B(H) - module generated by

{hj , j N}, i.e. HU is separable. Conversely, if HU is separable, there

exists a countable set {hj , j N} in HU that forms a gramian basis for

HU , wherefrom for each Dd , the Fourier development in HU of U

U () =

X

[U , hj ]hj (),

j=1

(2.4)

66

(iii) Now, if (2.2) holds, according to (ii) we have that HU is separable.

From the operatorial model of the normal Hilbert B(H) - module (see [6,

Corr.7, pp.30]) HU can be written as in (2.2) from [3] as the Hilbert space

tensor product H GU , hence the measurements space GU ( L20 ()) is

separable. Choosing an orthonormal basis {fj } in GU and h H, khk = 1,

it results (by Proposition 11, pp. 31 [6]) that the family {hfj }jN (where

(hfj )() is identified with the product fj h) is a gramian basis in H GU .

Thus it holds

X

U () =

[U , fj h]fj ()h,

(2.5)

j=1

from (2.5) we obtain (2.3).

Conversely, say that U has a representation (2.3), which means that U

is the limit of some linear combinations of elementary tensors from the

Hilbert space tensor product Sp {uj (), j N} Sp {fj , j N}.

Now, since for any Dd , we have

U Sp {uj (), j N} Sp {fj , j N} H Sp {fj , j N},

it results that HU H Sp {fj , j N}. The second member of the tensor

product being separable, it results that HU is separable.

Remark 2.4 If (, A , ) is separable, then each U Dd0 (H ) has all

types of series representation.

Corollary 2.5 Any stationary m.s.o.r.d.f. U has all types of series representation.

In the study of a m.s.o.r.d.f. U , its operator and scalar covariance kernels

U and U respectively, as well as it reproducing kernel normal B(H)module HU and reproducing kernel Hilbert space GU , respectively the

reproducing kernel Hilbert space KU play important roles. In what follows they will be connected to the series representations of U . First we easily deduce a series representation for the covariance kernels of a m.s.o.r.d.f.

U.

Proposition 3.1 If U is a m.s.o.r.d.f. of scalar covariance U and operator covariance U , then the following assertion hold:

(i) if U has a series representation (2.1), where hj is an orthonormal

system, then

U (, ) =

X

j=1

uj ()uj () ;

, Dd ;

(3.1)

67

{hj }jN is a gramian orthonormal system, then

U (, ) =

uj ()[hj , hj ]uj () ;

, Dd ,

(3.2)

j=1

(iii) if U has a tensor series representation of the form (2.3), such that

{fj } is an orthonormal system in L20 (), then

U (, ) =

uj () uj () ;

, Dd ,

(3.3)

j=1

holds, where the elementary tensors are understood in their sesquilinear form.

Proof. Direct verification.

Now, regarding the connections

of the series representations of U with the reproducing kernel structures

associated to U and U , we need first

Proposition 3.2 Let U = {U }Dd be a m.s.o.r.d.f. with scalar covariance U and operator covariance U . KU will be the reproducing kernel

Hilbert space associated to U , while HU the reproducing kernel module

associated to U .

(i) If {h }I is an orthonormal system in the vector time domain H(U )

and u () = (h , U )H , I, Dd , then {u }I form an

orthonormal system in KU .

(ii) If {h }I form a gramian orthogonal basis for the modular time

domain HU and u () = [h , U ] for I, Dd , then the family

{u }I forms a gramian basis for HU .

Proof. Since (i) is rather easy, we shall prove only (ii).

n

P

Let H0 be the set of distributions of the form

ak U (k , ), where ak

k=1

!

n

n

X

X

ak (k , ) =

ak Uk .

T0

k=1

k=1

Thus defined, T0 commutes with the modular action and preserves the

gramian. Since H0 is dense in HU and the range of T0 is dense in HU , T0

can be uniquely extended to a gramian unitary operator T : HU HU .

For any I, h can be expressed as

h =

a,k U,k ,

k=1

"

#

X

u () = [h , U ]H =

a,l U,l , U

H

l=1

X

l=1

a,l (,l , ) = T

X

l=1

!

a,l U,l

= T 1 h .

68

gramian basis in HU .

Combining Proposition 2.2 with Proposition

3.1 (ii) we get

Corollary 3.3 Let U = {U }Dd be a m.s.o.r.d.f. having the operator

correlation = U . If it admits a modular series representation, i.e. of

the form (2.2), where {hj }

j=1 is a gramian basis of HU , then the family

{uj () }

j=1 forms a gramian basis of the reproducing kernel module H .

The next result establishes a connection to the measurements space.

Theorem 3.4 Let U = {U }Dd be a m.s.o.r.d.f. having the operator

covariance = U . If it admits a tensor series representation, i.e. of the

form (2.3), where {fj }

j=1 is an orthonormal basis in the measurements

space GU L20 (), then {uj ()}

j=1 form an orthonormal basis in the

reproducing kernel Hilbert space GU .

j=1 be an orthonormal basis in GU . Then {fj }j=1 is an

2

orthonormal set in L0 (), which according to Proposition 2.2 means that

the coefficients (H-valued distributions) uj are uniquely determined. On

the other side, we have that for any h H, khk = 1, {(h fj )()}j =

{fj ()h}j form a gramian basis in H GU = HU (cf. ), from which,

writing the

P Fourier development of U with respect to this gramian basis

U () = [U , fj h]fj h, which is just (2.3) with

Dd , j 1.

uj () = [U , fj h]h,

{aj,m }

m=1 B(H) and {j,m }m=1 Dd , such that

fj ()h =

j 1.

m=1

"

uj () = [U , fj h]h = U ,

#

X

aj,m Uj,m h

X

X

[U , Uj,m ]aj,m h =

(, j,m )aj,m h.

m

69

space GU we have uj GU , for any j 1 and moreover,

!

X

X

(uj , uk )GU =

U (, j,m )aj,m h,

U (, k,n )ak,n h

m

m,n

m,n

"

=

ak,n Uk,n ,

aj,m Uj,m h, h

h)H

0

= (fk , fj )L2 () .

0

{uk }

k=1

is an orthonormal set in GU .

P

From the Corollary 3.3 we get U (, ) =

uk () uk (), for ,

Hence,

k=1

k=1 is complete in

GU and is thus an orthonormal basis.

References

[1] S. Cambanis, E. Masry, On the representation of weakly continuous

stochastic processes, Inform. Sci. 3, (1971), 277 290

[2] D. Gaspar, P. Gaspar, An operatorial model for Hilbert B(X )modules, An. Univ. Vest Tim. Ser. Matem.-Inform. XLI, fasc. 1

(2003), 49 64

[3] P. Gaspar, L. Popa, Stochastic Mappings and Random Distribution Fields. A Correlation Approach, arXiv:1406.2996v1 [math.FA],

11 Jun 2014

[4] P. Gaspar, L. Popa, Stochastic Mappings and Random Distribution

Fields II. Stationarity, arXiv:1406.2996v1 [math.FA], 11 Jun 2014

[5] Yu. Kakihara, Series Representations and Karhunen Processes, J.

Multiv. Analysis, 25, (1988), 10 24

[6] Yu. Kakihara, Multidimensional stochastic processes, World Sci

Publ., NJ, 1996

lineare Methoden in der Wahrscheinlichkeitsrechnung, Ann. Acad. Sci. Fenn. Ser. A I Math., 37, (1947), 1

79

`ve, Probability Theory, van Nostrad, New York, 1955

[8] M. Loe

[9] E. Masry, Expansion of multivariate weakly stationary processes,

Inform. Sci. 2, (1970), 303 317

On quaternions

Mot Ghiocel

Popa Lorena

Abstract

In this paper we present a logical and methodical approach of introducing the complex numbers and the quaternions and to link these concepts

with the theory of matrices.

Mathematics Subject Classification: 12F05, 97H40

Keywords: algebraic structure, complex number, quaternion, matrix

We consider the set C as the set of all ordered pairs of the real numbers

(a, b) and we denote:

C = R R = {(a, b)|a, b R}.

Theorem 1.1 The set C endowed with the operations:

(a, b) + (c, d) = (a + c, b + d)

(a, b) (c, d) = (ac bd, ad + bc)

where (a, b), (c, d) C, forms an algebraic structure of commutative field,

called the field of complex numbers.

Theorem 1.2 The set C endowed with the operations:

(a, b) + (c, d) = (a + c, b + d)

(a, b) = (a, b)

where R, (a, b), (c, d) C, forms an algebraic structure of linear space,

called the space of complex numbers. (see [3], pp.42)

Remark 1.3 The field C contains the field of real numbers R as a subfield

and it has the element i C such that: i2 = 1, i = (0, 1).

ghiocel.mot@uav.ro Department of Mathematics and Computer Science, Aurel Vlaicu

University of Arad, Romania

lorena.popa@uav.ro Department of Mathematics and Computer Science, Aurel Vlaicu

University of Arad, Romania

70

On quaternions

71

Indeed,

i2 = (0, 1) (0, 1) = (0 0 1 1, 0 1 + 1 0) = (1, 0) = 1C .

Using the notations 1R = (1, 0) si i = (0, 1) and the theorems [1.1] and

[1.2], we obtain the algebraic form of complex numbers:

(a, b) = (a, 0) + (0, b) = a + bi.

Therefore:

C = {a + bi|a, b R}.

The following proposition gives us a matrix representation of the complex numbers:

Proposition 1.4 The set of matrix:

a

C1 = Ma,b =

b

b

a, b R

a

an isomorphic field with the field C.

Ma,b + Mc,d = Ma+c,b+d

Ma,b Mc,d = Macbd,ad+bc .

This result is immediately by direct calculation (see [1]).

Remark 1.5 For any Ma,b C1 , we have:

Ma,b = a M1,0 + b M0,1 = aI2 + bM0,1 .

The another extension of the set of real numbers is given by the set of

quaternions. This set of hypercomplex numbers was introduced by W.

Hamilton in 1843.

Theorem 2.1 The set

R4 = R R R R = {(a, b, c, d)|a, b, c, d R}

forms an algebraic structure of noncommutative field with respect to operations:

(a1 , b1 , c1 , d1 ) (a2 , b2 , c2 , d2 ) = (a1 a2 b1 b2 c1 c2 d1 d2 , a1 b2 + b1 a2 +

+ c1 d2 d1 c2 , a1 c2 + c1 a2 + d1 b2 b1 d2 , a1 d2 + d1 a2 + b1 c2 c1 b2 )

(2.1)

where (a1 , b1 , c1 , d1 ), (a2 , b2 , c2 , d2 ) R4 . (see [3], pp.25)

On quaternions

72

Definition 2.2 The field R4 which contains the field of real numbers R

as a subfield and which has an imaginary units i, j, k C such that

i2 = 1, j 2 = 1, k2 = 1, ij = k = ji, jk = i = kj, ki = j = ik,

(2.2)

is called the quaternions field and it is denoted by K[i, j, k] or K.

Theorem 2.3 The set K[i, j, k] form an algebraic structures of linear

space with the operations:

(a1 , b1 , c1 , d1 ) + (a2 , b2 , c2 , d2 ) = (a1 + a2 , b1 + b2 , c1 + c2 , d1 + d2 )

(a, b, c, d) = (a, b, c, d)

where R, (a, b, c, d), (a1 , b1 , c1 , d1 ), (a2 , b2 , c2 , d2 ) K[i, j, k]. (see [3],

pp.44)

Because, the imaginary units are identified with the elements from

K[i, j, k]: i = (0, 1, 0, 0), j = (0, 0, 1, 0), k = (0, 0, 0, 1), using the theorems

[2.1] and [2.3], we have:

(a, b, c, d) = (a, 0, 0, 0)+(0, b, 0, 0)+(0, 0, c, 0)+(0, 0, 0, d) = a+bi+cj +dk.

Therefore,

K = {a + bi + cj + dk|a, b, c, d R, i, j, k with prop.[2.2]}.

To give a matrix representation for the quaternions, similarly with

[1.4] for complex numbers, we denote:

a + bi

c + di

a, b, c, d R .

, C =

K1 [i, j, k] =

(c di) a bi

Theorem 2.4 K1 [i, j, k] with the operations of matrix addition and matrix multiplication, forms a field isomorphic with K.

Proof. Let us consider the mapping:

: K K1 , (a + bi + cj + dk) =

a + bi

(c di)

c + di

.

a bi

(2.3)

direct calculation from [2.1] and from usual rules of addition and multiplication for matrices.

Remark 2.5 Any element from K1 [i, j, k] can be written as:

= aI2 + bI + cJ + dK,

1 0

i

0

0

1

0

where I2 =

,I =

,J =

,K =

0 1

0 i

1 0

i

i

.

0

On quaternions

73

Also, if we denote:

K2 [i, j, k] =

d

b

a

c

d

d

a, b, c, d R

b

d

c

a

b

only a real numbers.

Theorem 2.6 K2 [i, j, k] with the operations of matrix addition and matrix multiplications, forms a field isomorphic with K.

Proof. Let us consider the mapping:

a

b

: K K2 , (a + bi + cj + dk) =

d

c

b

a

c

d

d

c

a

b

c

d

.

b

a

(2.4)

direct calculation from [2.1] and from usual rules of addition and multiplication of matrices.

Remark 2.7 Any element from K2 [i, j, k] can be written as:

a

b d

c

b

a c d

= aI4 + bI + cJ + dK,

d

c

a

b

c

d

b

a

where

1

0

I4 =

0

0

0

0

J=

0

1

0

1

0

0

0

0

1

0

0

0

1

0

0

1

0

0

0

0

1

0

0

0

0

0

, I = 1 0

,

0

0

0

0

1

1

0

0 1 0

1

0

0

1 0

0

0

0

1

, K = 0

.

1

0

0

0

0

0

0 1

0

0

Let us mention that W. Hamilton also introduced a vector representation for quaternions: q = a + bi + cj + dk K, admits the vectorial

writing:

q =s+

v = [s,

v ],

(2.5)

v = a i + b j + c k - vector from the euclidean

space R3 .

From the relation [2.5] we easily deduce that:

v = s (a i + b j + c k );

the norm of a quaternion is: q = s2 (aa + b2 + c2 + d2 );

On quaternions

74

q1 q2 = (s1 +

v 1 )(s2 +

v 2 ) = s1 s2

v 1

v 2 +s2

v 1 +s1

v 2 +

v 1

v2

(2.6)

Remark 2.8 The quaternions which satisfy the condition s = 0, are

called vector - quaternions or ternions and they represent the image of a

real vector in quadridimensional space with the base (1, i , j , k ). Thus,

q1 q2 =

v 1

v2+

v1

v2

The vectorial representation of quaternions is useful for the representation

of rotations, which is actually based on multiplication between a quaternion and a vector.

Conclusions

C

= {a + bi|a, b R}

{a + cj|a, c R}

C=

C

= {a + dk|a, d R}.

But when we extend the set of complex numbers to the set of quaternions,

commutativity is lost. We mention also that: R C

= C1 K1 .

References

[1] A. Casu, I. Goian, P. Srbu, Sisteme numerice, CECMI, Chisin

au,

2008.

[2] A. M

arcus, Numere complexe, cuaternioni si aplicatii, Junior Summer

University - UBB, 19 July 2011.

[3] G. Mot, L. Popa , Algebr

a liniar

a, Geometrie analitic

a si diferential

a,

2014.

[4] J. Vince, Quaternions for Computer Graphics, Springer-Verlag, New

York, 2011.

for discrete-time linear systems in Banach spaces

Nicolae Marian Seimeanu

Mihail Megan

Abstract

This paper presents three concepts of uniform polynomial dichotomy

for linear discrete-time systems in Banach spaces. We treat the case in

which the sequence of projections is strongly invariant for the discrete

system (the evolution operator associated to the systems satisfies a local

invertibility property of the kernels of the projections). We give characterizations of the presented concepts and present the connections between

them through counterexamples. The main result of this paper is a result of boundedness of the sequence of projections that give a dichotomy

property.

Mathematics Subject Classification: 34D09

Keywords: uniform polynomial dichotomy; uniform strong polynomial dichotomy;

uniform weak polynomial dichotomy; discrete linear system.

Introduction

The property of dichotomy for evolution operators plays a key role in the

qualitative study of asymptotics of dynamical systems. Among the behaviors that govern these systems, one that led to many interesting results

is the exponential dichotomy property (both in the discrete / continuous and uniform / nonuniform framework) - see for example [1], [5], [13],

[16], [17], [19], [20], [21], [22], [23],[24]. More general behaviors can also

be considered, for example the (generalized) (h, k)-dichotomies ([2], [3],

[7], [10],[14], [15]) and (h, k, , )-dichotomies ([25]). As a natural relaxation of the exponential growth / decay, one can consider polynomial

growth rates with (at most) polynomial nonuniformities as asymptotics

for dynamical systems. This idea was successfully embraced and very

interesting results concerning the polynomial stability, instability and dichotomy were obtained, which vary from characterizations ([11], [12]) to

the study of stable manifolds ([6], [7], [8], [9]) and robustness of the polynomial dichotomy property ([4]).

In this paper we define three concepts of uniform polynomial dichotomy

nicolae.seimeanu77@e-uvt.ro Department of Mathematics,

West University of

Timisoara, Romania

megan@math.uvt.ro Department of Mathematics, West University of Timi

soara, Romania

77

78

projections. We emphasize the connections between them and propose

an open problem regarding the weak behavior defined in this paper. Under the assumption of boundedness of the sequence of projections, we

prove that the three concepts merge. The main result of our paper gives

an upper bound to the dichotomic sequence of projections, under the assumption of uniform polynomial growth of the linear discrete-time system

in discussion.

Preliminaries

Let X be a Banach space and B(X) the Banach space of all bounded

linear operators on X. The norms on X and on B(X) will be denoted

by . The identity operator on X is denoted by I. We will denote by

= {(m, n) N2 m n}.

We consider the linear difference system

xn+1 = An xn ,

(A)

Definition 2.1 The discrete evolution operator associated to the system (A) is defined, for (m, n) , by:

An

m ={

Am1 . . . An ,

I,

if m > n

if m = n

(2.1)

p

p

Remark 2.2 It is obvious that An

m An = Am , for all (m, n), (n, p)

and every solution of (A) satisfies xm = An

x

m n for all (m, n) .

Definition 2.3 The discrete linear system (A) is said to have a uniform

polynomial growth if there exist M, > 0 such that

An

m M (

m+1

) ,

n+1

(m, n) .

sequence of projections if Pn Pn = Pn for all n N.

If P N B(X) is a sequence of projections, then the sequence Q

N B(X) defined by Qn = I Pn is also a sequence of projections, called

the complementary sequence of projections of P N B(X) .

Definition 2.5 We say that the sequence of projections P N B(X) is

bounded if there exist M 1 such that

Pn M,

n N.

79

invariant for the system (A) if for all n N we have that

An Pn = Pn+1 An .

strongly invariant for the system (A) if it is invariant for (A)

and for all n N, the restriction An Ker Pn Ker Pn+1 is an

isomorphism.

If the sequence of projections P N B(X) is invariant for the system

(A), then we will say that (A, P ) is a dichotomy pair.

Remark 3.2 If a sequence of projections P N B(X) is strongly invariant for (A) then it is also invariant for (A). The converse is not

generally true, as the below example shows it.

Example 3.3 Let X = R3 and define, for every n N, define An , Pn

R3 R3 by

x, n 1

An x =

(x1 , 0, x3 ),

n=0

(x1 , x2 , 0), n = 0

Pn x =

(x1 + x2 , 0, 0), n 1.

(x1 , x2 , x3 ) R3 ,

An (x1 , x2 , 0), n = 0

An Pn x =

An (x1 + x2 , 0, 0), n 1

(x1 , 0, 0), n = 0,

=

(x1 + x2 , 0, 0), n 1

Pn+1 x, n 1

Pn+1 An x =

Pn+1 (x1 , 0, x3 ),

(x1 , 0, 0), n = 0

=

(x1 + x2 , 0, 0), n 1

and

n=0

But P N B(X) is not strongly invariant for (A). Assuming the

contrary, there exists x = (x1 , x2 , x3 ) R3 such that A0 x = y, where y =

(1, 1, 0) Ker P1 . We henceforth get the contradiction (x1 , 0, x3 ) =

(1, 1, 0).

Remark 3.4 If the sequence of projections P is strongly invariant for the

system (A) then

(i) for every n N there is an isomorphism Bn from the Ker Pn+1 to

Ker Pn such that An Bn Qn+1 = Qn and Bn An Qn = Qn ;

n

(ii) for all (m, n) there is an isomorphism Bm

from Ker Pm to

n

n n

Ker Pn with An

B

Q

=

Q

and

B

A

Q

=

Qn for all (m, n)

m

m m m

m m n

.

N B(X) to be a sequence of projections which is strongly invariant for

(A).

80

polynomially dichotomic (u.s.p.d) if there exist N 1 and > 0 such

that for all (m, n) the following conditions hold:

(uspd1 )

(m + 1) An

m Pn N (n + 1) ;

(uspd2 )

n

(m + 1) Bm

Qm N (n + 1) .

Remark 3.6 If the pair (A, P ) is u.s.p.d with constants N, > 0 given

by Definition 3.5, then for all n N, max{Pn , Qn } N .

Definition 3.7 We say that the pair (A, P ) is uniformly polynomially dichotomic (u.p.d) if there exist N 1 and > 0 such that for all

(m, n, x) X the following properties hold:

(upd1 )

(upd2 )

(m + 1) An

m Pn x N (n + 1) Pn x;

n

(m + 1) Bm

Qm x N (n + 1) Qm x.

Remark 3.8 If the pair (A, P ) is u.p.d with constants N, > 0 given by

Definition 3.7, then (upd2 ) is equivalent with the following condition:

(m + 1) Qn x N (n + 1) An

m Qn x,

(m, n, x) X.

Proposition 3.9 If the dichotomy pair (A, P ) is u.s.p.d then it is also

u.p.d.

Proof. If (A, P ) us u.s.p.d, let N, > 0 be such that (uspd1 ) and (uspd2 )

hold. The conclusion follows from the fact that for all (m, n, x) X

one has that

(m + 1) An

m Pn x (m + 1) Am Pn Pn x N (n + 1) Pn x

n

n

(m + 1) Bm

Qm x (m + 1) Bm

Qm Qm x N (n + 1) Qm x.

it is also u.s.p.d.

Proof. Let M 1 be an upper-bound for P N B(X) and N, > 0 be

given by the u.p.d property. Let x X and (m, n) . It follows that

(m + 1) An

m Pn x N (n + 1) Pn x 2N M (n + 1) x

n

(m + 1) Bm

Qm x N (n + 1) Qm x 2N M (n + 1) x.

From the above relations it follows that (A, P ) is u.s.p.d with constants

2M N, > 0.

bounded.

Remark 3.12 The converse implication from the preceding proposition

is not generally valid, as the following example shows it.

81

R2 consider the sequence of projections P N B(X) given by Pn (x1 , x2 ) =

(x1 + nx2 , 0), for n N and (x1 , x2 ) R2 . Define, for every n N,

An =

n+1

n+2

Pn +

(I Pn+1 ).

n+2

n+1

An

m =

n+1

m+1

Pn +

(I Pm ),

m+1

n+1

(m, n) .

It can easily be seen that P N B(X) is strongly invariant for (A), with

n

n+1

n+1

An

m Pn x = m+1 Pn x and Bm Qm x = m+1 Qm x, for all (m, n, x)

2

R hence (A, P ) is u.p.d. Having in mind that for all n N, Pn (0, 1) =

(n, 0) = n, it follows that sup Pn = + hence (A, P ) is not u.s.p.d.

nN

Definition 3.14 We say that the pair (A, P ) is uniformly weakly polynomially dichotomic (u.w.p.d) if there exist constants N 1 and > 0

such that for all (m, n) the following hold:

(uwpd1 )

(uwpd1 )

(m + 1) An

m Pn N (n + 1) Pn ;

n

(m + 1) Bm

Qm N (n + 1) Qm .

Remark 3.15 If the dichotomy pair (A, P ) is u.p.d then it is also u.w.p.d.

Open problem. At this point we ask wether the implication u.w.p.d

u.p.d generally holds or not.

Proposition 3.16 If (A, P ) is u.w.p.d and P is bouded then (A, P ) is

u.s.p.d.

Proof. Let M 1 be an upper bound for P N B(X) and N, > 0 be

given by the u.w.p.d property. Then, for all (m, n, x) X it follows

that

(m + 1) An

m Pn N (n + 1) Pn 2M N (n + 1)

n

(m + 1) Bm

Qm N (n + 1) Qm 2M N (n + 1)

Remark 3.17 If (A, P ) is u.w.p.d then it does not necessarily follow that

it is also s.p.d, fact pointed out by the following example.

Example 3.18 Consider the system given in Example 3.13. Because

(A, P ) is u.p.d it is also u.w.p.d, but (A, P ) is not u.s.p.d.

Remark 3.19 The connections between the concepts of uniform polynomial dichotomies presented in this paper are given by the following diagram.

u.s.p.d

u.p.d

u.w.p.d

82

Remark 3.20 If the sequence of projections which give the three concepts

of dichotomy for the system (A) is bounded, then we have that

u.s.p.d u.p.d u.w.p.d

The main result of the paper is given by the following

Theorem 3.21 Consider the system (A) having a uniform polynomial

growth. If there exists a strongly invariant sequence of projections P N

B(X) such that the dichotomy pair (A, P ) is u.p.d then P N B(X) is

bounded.

Proof. Let M, > 0 be given by the uniform polynomial growth of (A)

and N, > 0 given by the u.p.d property of (A, P ). Let x X, n N be

fixed an m n. Consider the below estimations:

Pn x x m + 1

m + 1

(

) N(

) Pn x

N

n+1

n+1

Qn x m + 1

m + 1

(

) N(

) Pn x

N

n+1

n+1

m+1

n

n

An

) x.

m Pn x Am Qm x Am x M (

n+1

It follows that

Pn x (

1

m + 1 2

m+1

N(

) )M(

) x.

N

n+1

n+1

(3.1)

< N1 and consider m = [0 + 1](n + 1) 1 n,

0

where [] is the integer part function. Then (3.1) becomes

Pn x (

and by denoting K =

1

N [0 + 1]2 ) M ([0 + 1]) x

N

1

N

([0 +1])

N [0 +1]2

Pn x Kx,

(n, x) N X.

References

[1] M.-G. Babutia, T. Ceausu, N. M. Seimeanu, On Uniform Exponential Dichotomy of Evolution operators, Analele Universit

atii de Vest,

Timisoara, Seria Matematic

a - Informatic

a, L, 2 (2012), 3-14.

[2] M.-G. Babutia, M. I. Kovacs, M. L

ap

adat, M. Megan, Discrete (h, k)-dichotomy and remarks on the boundedness of the

projections, Journal of Operators, Vol. 2014, Article ID 196345,

http://dx.doi.org/10.1155/2014/196345.

[3] M.-G. Babutia, M. Megan, I.-L. Popa, On (h, k)-dichotomies for

nonautonomous difference equations in Banach spaces, International

Journal of Differential Equations, Vol. 2013 (2013), Arcile ID 761680,

7 pages.

[4] L. Barreira, M. Fang, C. Valls, J. Zhang, Robustness of nonuniform polynomial dichotomies for difference equations, Topol. Method.

Nonl. An. 37 (2011), 357-376.

[5] L. Barreira, D. Dragicevic, C. Valls, Strong and weak (Lp , Lq )admissibility, Bull. Sci. math. 138 (6) (2014), 721-741.

[6] A. J. G. Bento, C. M. Silva, Stable manifolds for nonuniform polynomial dichotomies, J. Funct. Anal. 257 (2009) 122-148.

[7] A. J. G. Bento, C. M. Silva, Generalized nonuniform dichotomies and

local stable manifolds, J. Dyn. Diff. Equat. (2013). 25:1139-1158.

[8] A. J. G. Bento, C. M. Silva, Stable manifolds for non-autonomous

equations with non-uniform polynomial dichotomies, Q. J. Math.

63(2) (2012), 275-308

[9] A. J. G. Bento, C. M. Silva, Nonuniform dichotomic behavior: Lipschitz invariant manifolds for ODEs, Bull. Sci. Math. 138 (1) (2014),

89-109.

[10] M. Megan, On (h, k)-dichotomy of evolution operators in Banach

spaces, Dynam. Systems Appl. 5 (1996) no. 2, 189-195.

[11] M. Megan, T. Ceausu, A. A. Minda, On Barreira-Valls Polynomial

Stability of Evolution Operators in Banach Spaces, Electron. J. Qual.

Theory Differ. Equ. 33(2011), 1-10.

[12] M. Megan, T. Ceausu, M. L. R

amneantu, Polynomial stability of evolution operators in Banach spaces, Opuscula Mathematica 31 (2011),

no. 2, 279-288.

[13] M. Megan, A. L. Sasu, B. Sasu, Discrete admissibility and exponential

dichotomy for evolution families, Discret. Contin. Dynam. Systems 9

(2003), 383-397.

[14] R. Naulin, M. Pinto, Dichotomies and asymptotic solutions of nonlinear differential systems, Nonlinear Anal. 23(7) (1994), 871- 882.

[15] M. Pinto , Conditional stability for piecewise continuous solutions of

differential equations, Differ. Equ. Dyn. Syst. 2 (3) (1994), 217-230.

[16] I.-L. Popa, M. Megan, T. Ceausu, Exponential dichotomies for linear

difference systems in Banach spaces, Appl. Anal. Discrete Math., 6

(2012), 140-155.

[17] I.-L. Popa, M. Megan, T. Ceausu, Nonuniform exponential dichotomies in terms of Lyapunov functions for noninvertible linear

discrete-time systems, The Scientific World Journal, vol. 2013 (2013),

Article ID 901026.

[18] O. Perron, Die stabilit

atsfrage bei Differenzialgleischungen, Math. Z.

32 (1930), 703728.

[19] B. Sasu, A. L. Sasu, On the dichotomic behavior of discrete dynamical

systems on the half-line, Discrete Contin. Dyn. Syst. 33 (2013), 30573084.

[20] A. L. Sasu, M. G. Babutia, B. Sasu, Admissibility and nonuniform

exponential dichotomy on the half-line. Bull. Sci. Math., 137 (2013),

466484.

83

[21] A. L. Sasu, B. Sasu, Exponential dichotomy on the real line and admissibility of function spaces, Integr. Equat. Oper. Th. 54 (2006),

113-130.

[22] B. Sasu, Uniform dichotomy and exponential dichotomy of evolution

families on the half-line, J. Math. Anal. Appl. 323 (2006), 1465-1478.

[23] B. Sasu, A. L. Sasu, Exponential dichotomy and (lp , lq )-admissibility

on the half-line, J. Math. Anal. Appl. 316 (2006), 397-408.

[24] C. Stoica, M. Megan, Asymptotic behaviors for evolution equations in

infinite dimensional space, Proceedings of the International Symposium - Research and Education in an Innovation Era, Section Mathematics, 3rd Edition, Aurel Vlaicu University of Arad Publishing

House (2010) 187-191.

[25] J. Zhang, X. Chang, J. Wang, Existence and robustness of nonuniform (h, k, , )-dichotomies for nonautonomous impulsive differential equations. J. Math. Anal. Appl., 400 (2013), 710723.

84

L. Sida

P. Gaspar

Abstract

In this paper we extend the T -convolution product of periodic distributions to vector valued T -periodic distributions.

Mathematics Subject Classification: 58A30, 60E05

Keywords: periodic functions, periodic distributions, vector valued T - periodic

distributions

Introduction

In this paper we will define the T -convolution product for vector valued T - periodic distributions and extend some of the main properties

of this product to vector valued T - periodic distributions. Having in

mind the convolution product for vector valued distributions introduced

by L. Schwartz [9], we will define the T -convolution product for vector

valued T - periodic distributions in analogy to the T -convolution product

introduced by A. H. Zemanian (see [10], chap. 11) for scalar T -periodic

distributions.

As was stated and exemplified by A. H. Zemanian in [10, chp. 11]

for the scalar case, since T - periodic distributions do not have compact

support, we cant have a proper definition of the convolution product for

this distributions (excepting the case when one of this distributions is

null). To overcome this A. H. Zemanian introduced a special kind of the

convolution product (see [10]), called the T -convolution product.

Definition 2.1 ([10, chp.11, pp.314]) A function f : Rd C is said

to be periodic if there exists T = (T1 , T2 , ..., Td ) Rd , Ti > 0, such that

(LT f )(t) = f (t), t Rd , where L , Rd stands for the translation

operator on Rd . T is called a period of f . The set of all periods of f is

lavinia sd@yahoo.com Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

pastorel.gaspar@uav.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

85

86

fundamental period of f .

Definition 2.2 ([10, chp.11, pp. 314]) A function : Rd C will be

called T -periodic test function, if it is periodic of period T and infinitely

smooth. The space of all such T -periodic test functions will be denoted by

DT (Rd ) or Dd,T .

We shall also need the basic well known spaces of test functions

from distributions theory (see [1], [8]): D(Rd ), S(Rd ), E(Rd ), B(Rd ), which

we shall briefly denote Dd , Sd , Ed and Bd .

A function (x) is said to be unitary if it is an element of Dd and there

exists T Rd , Ti > 0 for which

X

(LnT )(x) = 1

nZd

for all x Rd (see [10]). The space of unitary functions will be denoted

by Ud,T ,

P

Remark 2.3 For every Dd , the sum

(LnT )(t) is finite and,

nZd

P

P

since LT (

(LnT )) =

(LnT ), it defines a function from Dd,T .

nZd

nZd

Definition 2.4 ([9, chp. II, 2]) Let E be a Banach space. Any linear

and continuous operator

U : Dd E

is an E-valued distribution on Rd .

Definition 2.5 ([9]) A vector valued distribution U is periodic of period

T Rd , Ti > 0 if LT U = U .

We shall also use the basic spaces from distributions theory, namely

D0 (Rd , E) = Dd0 (E) the space of all E-valued distribution on Rd ,

0

(E) the space of E-valued periodic distribution of

DT0 (Rd , E) = Dd,T

d

period T R , Ti > 0,

E 0 (Rd , E) = Ed0 (E) the space of E-valued distribution with compact

support,

B0 (Rd , E) = Bd0 (E) the space of E-valued bounded distributions.

Remark 2.6 The derivative of an E-valued tempered distribution is also

an E-valued tempered distribution. The E-valued tempered distributions

generalize the bounded (slow-growing) locally integrable E-valued functions; all E-valued distributions with compact support and all squareintegrable E-valued functions are E-valued tempered distributions.

We also recall the inclusion relations (with continuous embedding)

between the spaces from distribution theory:

0

Dd,T

(E) Bd0 (E) Sd0 (E) Dd0 (E).

(2.1)

87

operator UF defined by

Z

UF () := (t)F (t)dt, Dd

(2.2)

Rd

Every E-valued tempered distribution is a derivative of finite order

of some continuous E-valued function of polynomial growth (the structure

theorem for Sd0 (E)).

Let us recall the function from L1d,T , and, by analogy with the case

of non-periodic functions from L1d , we will quote a theorem for the T convolution product for scalar T -periodic functions (meaning from L1d,T ).

We note first that for a function f L1d,T , T Rd , Ti > 0, for any

d

a R the following equality holds (see [6]):

Z

Z

f (t)dt =

f (t)dt.

(3.1)

[a,a+T ]

[0,T ]

L1d,T .

Theorem 3.1 Let f, g L1d,T . Then for any s [0, T ), the function

t f (t)g(s t) is in L1d,T . The T -convolution f T g defined by:

Z

(f T g)(s) =

f (t)g(s t)dt, s [0, T )

(3.2)

[0,T )

L1d,T

and also

k f T g k1 k f k1 k g k1 .

(3.3)

Thus L1T becomes a Banach algebra with the T -convolution for function

as multiplication.

The proof of this theorem is based on the reasoning presented in [1] paragraph 2.4.3, p. 129, and for the T -periodicity we have

Z

Z

g(tT )=g(t)

(f T g)(tT ) =

f (s)g(tT s)dt =

f (s)g(tsT )dt

=

[0,T ]

g(tT )=g(t)

[0,T ]

[0,T ]

only one is periodic, is a periodic function of same period, i.e. if f Dd,T

si g Dd , then f g Dd,T . (see [6], p. 14)

88

Proposition 3.3 (see [6]) The T -convolution product f T g is a continuous application in the meaning of the topology from L1d,T , i.e. for any

convergent sequence (fn )n1 from L1d,T , we have

L1

T

fn T g

f T g for any g L1T .

At the beginning let us recall same known results for the convolution product of scalar distributions.

The convolution between a distribution u and a function Dd :

(u )(t) = (u, t ), Dd , t Rd

(3.4)

where t = Lt .

etc. (see [1], p. 212)

The convolution product between two distributions, one of them

with compact support u Dd0 , v Ed0 :

(u v, ) = (u, v ), Dd .

(3.5)

etc.(see [1], p. 216)

If u Dd0 , v Ed0 then u v Dd0 , or if u Ed0 , v Ed0 then u v Ed0 , etc.

The T -convolution between two T -periodic distributions uT , vT

(Dd,T )0 (or finite convolution):

(uT T vT , ) = vT (

uT T ), Dd,T ,

(3.6)

The T -convolution between a T -periodic distributions uT

(Dd,T )0 and a test T -periodic function Dd,T :

= (u, t ), t Rd , Ud,T .

(uT T )(t) = uT (Lt )

(3.7)

For the case of vector valued distributions we consider the locally

convex quasi - complete space X. Let U Dd0 (X) and V Ed0 (X).

Definition 3.4 The convolution product between two vector valued distributions U Dd0 (X), V Ed0 (X), is defined as:

), Dd ,

(U V )() = V (U

(3.8)

We mention that, the above definition obviously makes sense if U

is a scalar distribution. In this general case the formula (3.8) is meant in

the weak sense, i.e.

) , x0 X 0 .

x0 (U V ()) = V x0 (U

(3.9)

We recall here some properties of the convolution product between

two vector valued distributions:

89

U = U, U Dd0 (X).

(3.10)

D U = U D , N.

(3.11)

From the convolution with a test function Dd results the regularization of a vector valued distribution by a scalar function, infinitely

differentiable. (see [9], vol. II p. 72)

(U )(t) = (Lt U )(),

D d , t Rd .

We mention that, the convolution of a vectorial distribution with

a scalar function c

a, in particular, if U a temperate distribution, then

the convolution have sense for any Sd , and if U is a distribution with

compact support then the convolution have sense for any Ed .

In the next definition we extend the notion of T -convolution product, for the case of vector valued T -periodic distributions.

Definition 3.5 The T -convolution product between two vector valued T periodic distributions UT , VT B(Dd,T , X) is:

T T ),

(UT T VT )() = VT (U

(3.12)

We mention that

T T )(t) = (U

T , Lt ) = (UT , (Lt ) ) = (U, (Lt ) ),

(U

0

.

like function of t, is (see [10] Cor. 2.7-2a, p. 74) from Dd,T

Remark 3.6 Because (U

relation (3.12) always exist for any choice of UT si VT from B(Dd,T , X).

This is in contrast with the usual convolution product of two distributions

of Dd0 (X), where, to ensure the existence of convolution, it requires to

introduction the additional restrictions.

Remark 3.7 The definition (3.12) of T -convolution product is correct,

since for regular periodic distributions it coincides with the standard definition of the T -convolution of periodic functions.

In the next theorem we show that, the space of vector valued T periodic distributions is closed under the T -convolution product.

Theorem 3.8 The T -convolution product between two vector valued T periodic distributions

B(Dd,T , X) is still in B(Dd,T , X). In other words, the space B(Dd,T , X)

is closed under the T -convolution product.

Proof. We show that (3.12) defines an operator from B(Dd,T , X).

For the linearity of UT T VT we consider , C, and 1 , 2

Dd,T . Then from the relation (3.12) it results

T T (1 + 2 )) =

(UT T VT )(1 + 2 ) = VT (U

90

T T 1 ) + (U

T T 2 )) =

= VT ((U

T T 1 ) + VT (U

T T 2 ) =

= VT (U

= (UT T VT )(1 ) + (UT T VT )(2 ).

For the continuity UT T VT we consider the sequence {k }

k=1

in Dd,T . Now

, (t)(s)k (t + s)) (V U

, (t)(s)(t + s)) =

(UT T VT )(k ) = (V U

= (UT T VT )(),

because if {k }

n Dd as well.

k=1 in Dd,T then {k }k=1

So, UT T VT B(Dd,T , X), where it results that UT T VT = WT

0

where W Dd,T

(X), i.e. W is T -periodic.

References

[1] D. Gaspar, P. Gaspar, Analiz

a functional

a, Ed. a 2-a, Ed. de Vest,

Timisoara, 2009.

[2] P. Gaspar, L. Sida: Periodically correlated multivariate second order

random distribution fields and stationary cross correlatedness, J. Func.

Anal.(2014), Volume 267, Issue 7, 1 October 2014, Pages 2253-2263,

ISSN 0022-1236, http://dx.doi.org/10.1016/j.jfa.2014.07.019.

da

ban, L. Sida, On vector valued periodic distribu[3] P. Gaspar, S. Na

tions, Theory and Applications of Mathematics & Computer Science

no. 2 (1), (2012), 1 - 9.

[4] Y. Kakihara, Multidimensional Second Order Stochastic Processes,

World Scientific Publ.Comp., River Edge, N.I., 1997.

[5] W. Kecs, Produsul de convolutie si aplicatii, Editura Academiei RSR,

Bucuresti, 1978.

, Teoria distributiilor, functii complexe si

[6] V. olariu, V. Prepelit

a

aplicatii , Editura S

tiintific

a si enciclopedic

a, Bucuresti, 1986.

[7] L. Popa, On vector valued distributions, Proceedings of the International Symposium Research and Education in an Innovation

era,Arad, 16 - 18 Nov. 2006, Section Mathematics and Computer

Science, (S. N

ad

aban and C. Stoica edtrs.), Ed. Mirton, Timisoara,(

2007), 112 117.

[8] L. Schwartz, Theorie des distributions I, II, Fasc. IX, X, Actualites

Scientifiques Industrielles 1245, 1112, Hermann, Paris, 1957, 1959.

[9] L. Schwartz, Theorie des distributions a

valeurs vectoriells I, II,

Ann. Inst. Fourier, VII (1957), VIII (1958), 1 141, 1 209.

[10] A. H. Zemanian, Distributions theory and transform analysis, New

York - Saint Louis -San Francisco - Toronto - London - Sydney, Mc

Graw-Hill Book Co., Inc., (1965).

Sandru Andrea

Palcu Adrian

Abstract

Chemical Graph theory is used to model physical properties of

molecules called alkanes. Indices based on the graphical structure of the

alkanes are defined and used to model both the boiling point and melting

point of the molecules.

Mathematics Subject Classification: 94C15, 97K30

Keywords: Chemical Graph theory, alkanes, boiling point

Introduction

This article has arisen because the curriculum of computer science states

that the examples used in teaching will be mainly drawn from the curriculum Mathematics and Natural Sciences. Because in class XI studying

concept graph and his applications, we thought it would be a good example linking the concepts learned in chemistry with concepts learned

in computer science. Chemical graph theory is a branch of mathematics

which combines graph theory and chemistry. Graph theory is used to

mathematically model molecules in order to gain insight into the physical

properties of these chemical compounds. Some physical properties, such

as the boiling point, are related to the geometric structure of the compound. This is especially true in the case of chemical compounds known

as alkanes. Alkanes are organic compounds exclusively composed of carbon and hydrogen atoms. One example of an alkane is ethane, shown in

Figure 1.

sandru.andrea@uav.ro Department of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

palcu.adrian@uav.ro DDepartment of Mathematics and Computer Science, Aurel

Vlaicu University of Arad, Romania

91

92

Figure 1:

Each carbon atom has four chemical bonds and each hydrogen atom has

one chemical bond. Therefore, the hydrogen atoms can be removed without losing information about the molecule. The resulting representation

of ethane is the carbon tree shown in Figure 2.

Figure 2:

This carbon tree can be represented as a graph by replacing the carbon

atoms with vertices. Chemical bonds are then represented as an edge in

the graph. Figure 3 shows the graphical representation of ethane composed of two vertices connected by a single edge.

Figure 3:

Main results

properties of alkanes can be modeled using topological indices. Some

of these indices are well known outside of the chemical and mathematical

communities such as the relative molecular mass (Mr) of a compound. For

alkanes, the relative molecular mass is a function of the number of carbon

atoms, denoted by n, and is given by Mr(n) = 12.01115n + 1.00797(2n +

2) atomic mass units (amu). Using this formula, you can determine that

the relative molecular mass of ethane in Figure 3 is 30.0701amu.

The boiling point of alkanes is determined by the geometric structure of

the alkane. Boiling points are a measure of the forces of attraction between

like molecules. For essentially nonpolar compounds such as alkanes, these

forces are London dispersion forces due to instantaneous dipole-induced

93

dipole attractions. Dispersion forces are very short range forces which increase with the number of electrons which is proportional to the relative

molecular mass for the alkanes. The alkane boiling point should depend

on the relative molecular mass and on how well the molecules pack together, which is related to the geometry of the molecule. The dependence

on the geometry is complex, but the boiling point should decrease in a

general way as the compactness of the molecule increases if the relative

molecular mass stays the same. Balaban noted that for the same relative

molecular mass, the boiling point decreased with increased branching.

We examine the structures and boiling points of octane and

2,2,4-trimethylpentane to illustrate this result. Both are composed of

8 carbon atoms so they have the same molecular weight. A 3D representation is given in Figure 4 and the graphical representation is given in

Figure 5.

Figure 4:

Figure 5:

2,2,4-trimethylpentane is a more compact alkane and is sometimes called

isooctane. A 3D representation is given in Figure 6 and the graphical

representation is given in Figure 7.

From the above discussion, we expect the boiling point of

2,2,4-trimethylpentane to be lower than that of octane. This is indeed

the case. The boiling point of octane is 398.7 K while the boiling point of

2,2,4-trimethylpentane is 372.4 K. It is possible to model the boiling point

of families of alkanes having similar geometric structure using molecular

weight as the only index in the model. In modeling the alkanes in general, more topological indices are needed to reduce the error in the model.

Some examples include the Hosoya index, the Wiener number, the Wiener

path numbers, the Mean Wiener index, and the Methyl index.

The Hosoya index (denoted Z) is the sum of the coefficients of the simple

matching polynomial for a graph. This is equivalent to the number of

matchings a graph contains plus 1 to account for the matching consisting

of no edges. A matching of a graph G is a (possibly empty) set of edges

94

Figure 6:

Figure 7:

of G in which no two edges share a common vertex. The set of edges in

a matching are said to be independent. An algorithm for computing the

simple matching polynomial of a graph is given by Farrell. The Hosoya

index for ethane is 2 since it contains only one edge yielding one matching

with zero edges and one matching with one edge. Using this algorithm

we determine the Hosoya index for 2,2,4-trimethylpentane is 19. You can

verify this index by carefully inspecting all the sets of independent edges.

There is 1 way to choose zero edges, there are 7 ways to choose only one

edge in the matching, 11 ways to choose two edges in the matching, and

there is no way to choose three or more edges for a matching. This gives,

1 + 7 + 11 = 19 simple matchings of 2,2,4-trimethylpentane verifying the

Hosoya index for this alkane.

The Wiener number (denoted W) is the sum of the distances between all

pairs of vertices in a graph. It can be computed by adding the entries

in the upper (or lower) triangular part of the distance matrix of a graph.

Ethane has a Wiener number of 1 since it has only one pair of vertices

separated by an edge. For 2,2,4-trimethylpentane, we use the distance

matrix in Figure 9 computed from the labeled graph in Figure 8. The

Wiener number of 2,2,4-trimethylpentane is therefore 66.

The Wiener path numbers (denoted 1 P , 2 P , 3 P , 4 P , ...) are defined by

i

P which is the number of pairs of vertices in the graph separated by i

edges. i P can be computed using the distance matrix of a graph and

counting the number of times i appears in the upper triangular part of

95

Figure 8:

Figure 9:

the matrix. Using the distance matrix for 2,2,4-trimethylpentane given

in Figure 9, we find that 1 P = 7, 2 P = 10, 3 P = 5, and 4 P = 6. The

Mean Wiener index (denoted W) is the average of the distances between

all pairs of vertices in a graph. We previously showed the Wiener number for 2,2,4-trimethylpentane is 66. Using this, we calculate that W =

66/28 = 2.35714. The methyl index was introduced to help graphically

represent the branching of the alkanes. The methyl index (denoted Mth)

is defined to be the number of degree one vertices which are adjacent to a

vertex of degree three or greater. For example, the methyl index for 2,2,4trimethylpentane is 5. The five methyl edges as seen in Figure 8 are (1, 2),

(2, 6), (2, 7), (4, 5), and (4, 8). All of the indices described can be used to

construct models of the various physical properties of alkanes. Several of

these indices were used to model the boiling points of alkanes having six to

twelve carbon atoms. The normal alkanes with thirteen through twentytwo carbons were also included to facilitate the prediction of test data

having thirteen to twenty-two carbons. The total number of alkanes modeled was 187. One such model is f (1 P,2 P, ...,6 P, M th, Z) = 847.41474 +

221.61698(1 P )0.49420 1182.20853(2 P )0.03689 + 0.00125(3 P )3.39724

3.02445(4 P )0.93751 2.16070(5 P )1.01631 0.56366(6 P )1.38233

2.10575M th0.5695 9.61075Z 0.19907

The coefficient of determination for this model is 0.997068 and the standard deviation is 2.1 degrees (C). Table 1 gives the number and percentage

of alkanes with the specified absolute boiling point deviations given by this

96

model.

BP dev.

0 10

1 20

2 40

4 60

6 90

> 90

alkanes

84

48

46

5

3

1

% alkanes

44.9

25.7

24.6

2.7

1.6

0.5

References

[1] Balaban, Alexandru T. (2001). A Personal View About Topological

Indices for QSAR / QSPR, QSPR / QSAR Studies by Molecular Descriptors, Nova Science Publishers, Huntington, NY, 2001, 1-30.

[2] Burch, Kimberly J., Wakefield, Diane K., and Whitehead, E. Glen, Jr.

(2003). Boiling Point Models of Alkanes, MATCH Communications

in Mathematical and Computer Chemistry 47, 25-52.

[3] Farrell, Edward J. (1979). An introduction to matching polynomials,

J. Combinat. Theory Ser. B 27, 75-86.

[4] Kimberly Jordan Burch (Indiana Univ. of Pennsylvania), Chemical

Graph Theory, Loci (February 2010), DOI:10.4169/loci002857

- Pseudo Lattice Graphs and their Applications to Fuzzy and Neutrosophic ModelsDiunggah olehAnonymous 0U9j6BLllB
- Collected Papers, Vol. 5Diunggah olehAnonymous 0U9j6BLllB
- Arabic English GlossaryDiunggah olehishaq2628
- FlorentinSmarandacheDiunggah olehAnonymous 0U9j6BLllB
- Data Hiding in Audio FilesDiunggah olehgdskumar
- New Techniques to Analyse the Prediction of Fuzzy ModelsDiunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Theory and its Applications, Collected Papers, Vol. 1Diunggah olehAnonymous 0U9j6BLllB
- New Research on Neutrosophic Algebraic StructuresDiunggah olehAnonymous 0U9j6BLllB
- Complete Synchronization Of Hyperchaotic Xu And Hyperchaotic Lu Systems Via Active ControlDiunggah olehijcses
- Multi-matrices and arithmetical operations with multi-matrices, by Constantin ScheauDiunggah olehFlorentin Smarandache
- Solving Diophantine EquationsDiunggah olehAnonymous 0U9j6BLllB
- Progress in Physics, Vol. 1, 2015Diunggah olehAnonymous 0U9j6BLllB
- Scientia Magna, Vol. 10, No. 1, 2014Diunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Sets and Systems, Vol.-6, -2014Diunggah olehAnonymous 0U9j6BLllB
- International Journal of Mathematical Combinatorics, Vol. 4, 2014Diunggah olehAnonymous 0U9j6BLllB
- An Introduction to Dynamical Systems and ChaosDiunggah olehLinaena Mericy
- Correspondence PrincipleDiunggah olehGum Dropz
- LaSalle’s Invariance Principle & Chetaev’s Theorem copyDiunggah olehMutt3012
- 07 Witrant SpringerDiunggah olehRizaldy Pahlevi Bin Jamaluddin
- C0603021013.pdfDiunggah olehtheijes
- Gandolfo Cap 15 EdDiunggah olehMiguel Perez Elizalde
- Nonlinear Control And Servo SystemsDiunggah olehshivnair
- SteganographyDiunggah olehshelaranchashrikant
- MIT6_241JS11_chap14Diunggah olehAntonio Perez
- 33204_06_SysDiffEquDiunggah olehVinicius
- PSMobiusDiunggah olehamra_a86
- CS421_tutorial.pdfDiunggah olehMtanzania Mtanzania
- Analysis of Numerical SchemesDiunggah olehNishant Manepalli
- rizzi unit plan - exponential functionsDiunggah olehapi-241764369
- ForteGiroscopice_evitarea_coliziuniiDiunggah olehAna Maria Ponea

- TOPSIS Method for MADM based on Interval Trapezoidal Neutrosophic NumberDiunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Sets and Systems. Book series, vol. 23 / 2018Diunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Sets and Systems. Book series, vol. 22 / 2018Diunggah olehAnonymous 0U9j6BLllB
- VIKOR based MAGDM Strategy with Trapezoidal Neutrosophic NumbersDiunggah olehAnonymous 0U9j6BLllB
- Operators on Single Valued Trapezoidal Neutrosophic Numbers and SVTN-Group Decision MakingDiunggah olehAnonymous 0U9j6BLllB
- Validation of the pedagogical strategy for the formation of the competence entrepreneurshipin high education through the use of neutrosophic logic and Iadov techniqueDiunggah olehAnonymous 0U9j6BLllB
- The Basic Notions for (over, off, under) Neutrosophic Geometric Programming ProblemsDiunggah olehAnonymous 0U9j6BLllB
- Some Neutrosophic Probability DistributionsDiunggah olehAnonymous 0U9j6BLllB
- Optimization of EOQ Model with Limited Storage Capacity by Neutrosophic Geometric ProgrammingDiunggah olehAnonymous 0U9j6BLllB
- Extension of Soft Set to Hypersoft Set, and then to Plithogenic Hypersoft SetDiunggah olehAnonymous 0U9j6BLllB
- Single Valued Neutrosophic Hesitant Fuzzy Computational Algorithm for Multiobjective Nonlinear Optimization ProblemDiunggah olehAnonymous 0U9j6BLllB
- Extensions to Linguistic Summaries Indicators based on Neutrosophic TheoryDiunggah olehAnonymous 0U9j6BLllB
- A New Neutrosophic Cognitive Map with Neutrosophic Sets on ConnectionsDiunggah olehAnonymous 0U9j6BLllB
- On Single Valued Neutrosophic Signed Digraph and its applicationsDiunggah olehAnonymous 0U9j6BLllB
- A framework for selecting cloud computing services based on consensus under single valued neutrosophic numbersDiunggah olehAnonymous 0U9j6BLllB
- Gaussian single-valued neutrosophic numbers and its application in multi-attribute decision makingDiunggah olehAnonymous 0U9j6BLllB
- On single-valued co-neutrosophic graphsDiunggah olehAnonymous 0U9j6BLllB
- Sinos River basin social-environmental prospective assessment of water quality management using fuzzy cogmitive maps and neutrosophic AHP-TOPSISDiunggah olehAnonymous 0U9j6BLllB
- Positive implicative BMBJ-neutrosophic ideals in BCK-algebrasDiunggah olehAnonymous 0U9j6BLllB
- Performance Evaluation of Mutual Funds Via Single Valued Neutrosophic Set (SVNS) Perspective: A Case Study in TurkeyDiunggah olehAnonymous 0U9j6BLllB
- On Neutrosophic Crisp Topology via N-TopologyDiunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Soft Normed Linear SpacesDiunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Soft Cubic Set in Topological SpacesDiunggah olehAnonymous 0U9j6BLllB
- Neutrosophic Shortest Path ProblemDiunggah olehAnonymous 0U9j6BLllB
- MBJ-neutrosophic structures and its applications in BCK/BCI-algebrasDiunggah olehAnonymous 0U9j6BLllB
- Intrusion Detection System and Neutrosophic Theory for MANETs: A Comparative StudyDiunggah olehAnonymous 0U9j6BLllB
- A new approach for multi-attribute decision-making problems in bipolar neutrosophic setsDiunggah olehAnonymous 0U9j6BLllB
- Algebraic Structure of Neutrosophic Duplets in Neutrosophic RingsDiunggah olehAnonymous 0U9j6BLllB

- activity portfolio - lyricsDiunggah olehapi-301527166
- Aqa Gcse Science Command WordsDiunggah olehan7li721
- Gantt Charts.docxDiunggah olehVikas Sharma
- What is Linguistics?PDFDiunggah olehGlennmelove
- Review on Hybrid LUT/Multiplexer Combinational Logic Block ArchitectureDiunggah olehIJRT Online
- 07 Conductivity Solutions TgDiunggah olehrogue
- Meamber, Laurie A. “Postmodernism and Arts Marketing.” In The Routledge Companion to Arts Marketing. Eds. O’Reilly, D. Rentschler, R. & Kirchn ....pdfDiunggah olehFelippe Fritz Delangelo
- DementiaDiunggah olehhhamiem
- PersuasiveDiunggah olehdeepsjyoti
- bloomDiunggah olehJaycelyn Alejo Grospe
- Value Partners Q4 2014 CommentaryDiunggah olehCanadianValue
- BatteriesDiunggah olehcmscosta
- Material Sheet Solution Ch 4Diunggah olehDavid Sahry Mark
- Masin_MPhil ThesisDiunggah olehrakesh0013
- Shakespeare - MacbethDiunggah olehAshwin Anuraag
- Blue Ocean Strategy Project ReportDiunggah oleharpit vora
- Chapter 1 - Collaborative Advantage What Why How and Why NotDiunggah olehAmethyst Online
- The origins of Anglo-Saxon kingdoms.pdfDiunggah olehEric Dubourg
- An Eye Witness Account of the Descent and Ascent of Lord Shiva From Mt. KailashDiunggah olehWillard Van De Bogart
- Caelan Barr Creative Portfolio 2014Diunggah olehRudy Cox
- The Antioxidant MythDiunggah olehcygnus_11
- 3137503 Literature Review Article Electroinc Medical Records Stimulation for Nursing Students 3 5Diunggah olehAndrea Tyler
- Session 5 Culture of RespectDiunggah olehRohan Shah
- DECISION SUPPORT SYSTEM FOR MATCHING TRACTOR POWER AND IMPLEMENT SIZE IN IRRIGATED FARMING OF SUDANDiunggah olehInternational Educational Applied Scientific Research Journal (IEASRJ)
- Written ReportDiunggah olehCamille Lucelo
- Ingles - Ultimo TesteDiunggah olehDiogo Ramalho
- Jurnal readingDiunggah olehNurul Mazni
- outcome 2- science 9 unit planDiunggah olehapi-310123009
- Ephraim (Moraitis) of Philotheou - OrthodoxWikiDiunggah olehSancta Maria Servus
- Target class guided feature subsettingDiunggah olehmeenakshi_scribd